Spaces:
Running
Running
Download index.html from IntelligenceLab/AI-for-Science: direct link, hf CLI and curl.
- Browser
- Download file 55.7 kB
-
https://huggingface.co/spaces/IntelligenceLab/AI-for-Science/resolve/main/index.html
- Command line
-
hf download hf://spaces/IntelligenceLab/AI-for-Science/index.html
-
curl -L -o index.html https://huggingface.co/spaces/IntelligenceLab/AI-for-Science/resolve/main/index.html
55.7 kB
| <html lang="en"> | |
| <head> | |
| <meta charset="utf-8"> | |
| <meta name="viewport" content="width=device-width, initial-scale=1"> | |
| <title>AI for Science Collaboration · IntelligenceLab</title> | |
| <meta name="description" content="We're looking for biology and materials science graduate students to collaborate on AI for Science: tell us how you do research, help us test AI agents on real scientific workflows, and co-author the results."> | |
| <meta property="og:type" content="website"> | |
| <meta property="og:title" content="AI for Science Collaboration · IntelligenceLab"> | |
| <meta property="og:description" content="Help us build AI agents that actually help biology and materials science research, and write the paper with us."> | |
| <meta name="theme-color" content="#faf8f3"> | |
| <link rel="icon" type="image/svg+xml" href="data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 32 32'%3E%3Cdefs%3E%3ClinearGradient id='g' x1='0' y1='0' x2='1' y2='1'%3E%3Cstop offset='0' stop-color='%2310b981'/%3E%3Cstop offset='1' stop-color='%236366f1'/%3E%3C/linearGradient%3E%3C/defs%3E%3Crect width='32' height='32' rx='8' fill='url(%23g)'/%3E%3Cpath d='M7 23h5v-5h5v-5h5' fill='none' stroke='white' stroke-width='2.6' stroke-linecap='round' stroke-linejoin='round'/%3E%3Cpath d='M19.6 9.4l2.2 2.2 4.2-4.4' fill='none' stroke='white' stroke-width='2.6' stroke-linecap='round' stroke-linejoin='round'/%3E%3C/svg%3E"> | |
| <link rel="preconnect" href="https://fonts.googleapis.com"> | |
| <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin> | |
| <link href="https://fonts.googleapis.com/css2?family=Fraunces:ital,opsz,wght@0,9..144,500;0,9..144,600;1,9..144,500&family=Inter:wght@400;500;600;700&family=JetBrains+Mono:wght@400;500;600&display=swap" rel="stylesheet"> | |
| <link rel="stylesheet" href="styles.css"> | |
| <script>document.documentElement.classList.add('js');</script> | |
| </head> | |
| <body> | |
| <a class="skip-link" href="#main">Skip to content</a> | |
| <header class="nav"> | |
| <div class="container nav-inner"> | |
| <a class="brand" href="#top" aria-label="IntelligenceLab AI for Science home"> | |
| <svg class="brand-mark" viewBox="0 0 32 32" aria-hidden="true"> | |
| <defs> | |
| <linearGradient id="brand-grad" x1="0" y1="0" x2="1" y2="1"> | |
| <stop offset="0" stop-color="#10b981"/> | |
| <stop offset="1" stop-color="#6366f1"/> | |
| </linearGradient> | |
| </defs> | |
| <rect width="32" height="32" rx="8" fill="url(#brand-grad)"/> | |
| <path d="M7 23h5v-5h5v-5h5" fill="none" stroke="#fff" stroke-width="2.6" stroke-linecap="round" stroke-linejoin="round"/> | |
| <path d="M19.6 9.4l2.2 2.2 4.2-4.4" fill="none" stroke="#fff" stroke-width="2.6" stroke-linecap="round" stroke-linejoin="round"/> | |
| </svg> | |
| <span>IntelligenceLab <em>AI for Science</em></span> | |
| </a> | |
| <nav class="nav-links" aria-label="Primary"> | |
| <a href="#why">Why</a> | |
| <a href="#your-research">Your research</a> | |
| <a href="#examples">Demos</a> | |
| <a href="#collaborate">How we work</a> | |
| <a href="#papers">Papers</a> | |
| </nav> | |
| <div class="nav-cta"> | |
| <a class="btn btn-dark btn-sm" href="#join">Get involved</a> | |
| <button class="nav-toggle" type="button" aria-label="Open menu" aria-expanded="false" aria-controls="mobile-menu"> | |
| <span></span><span></span> | |
| </button> | |
| </div> | |
| </div> | |
| <div class="mobile-menu" id="mobile-menu" hidden> | |
| <a href="#why">Why</a> | |
| <a href="#your-research">Your research</a> | |
| <a href="#examples">Demos</a> | |
| <a href="#collaborate">How we work</a> | |
| <a href="#papers">Papers</a> | |
| <a class="btn btn-dark" href="#join">Get involved</a> | |
| </div> | |
| </header> | |
| <main id="main"> | |
| <section class="hero" id="top" aria-labelledby="hero-title"> | |
| <div class="container hero-grid"> | |
| <div class="hero-copy"> | |
| <p class="eyebrow"><span class="pulse-dot" aria-hidden="true"></span>Open call · Biology & materials science</p> | |
| <h1 id="hero-title">Building AI that actually <em>helps your research</em>.</h1> | |
| <p class="lede">We're AI researchers who teach language-model agents to use real scientific software. We already have research directions and tasks in mind. What we need is your expertise: propose the verifiers that check whether an agent's result is scientifically right, and help us decide how its work should be evaluated. If you're a graduate student in biology or materials science, come design them with us and write the paper together.</p> | |
| <div class="hero-actions"> | |
| <a class="btn btn-dark" href="#join">Join the collaboration | |
| <svg viewBox="0 0 20 20" aria-hidden="true"><path d="M4 10h11m-4-4 4 4-4 4" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round"/></svg> | |
| </a> | |
| <a class="btn btn-ghost" href="#examples">See our demos</a> | |
| </div> | |
| <ul class="hero-points"> | |
| <li>No ML background needed</li> | |
| <li>Your workflows, your expertise</li> | |
| <li>Contributors can join the paper</li> | |
| </ul> | |
| </div> | |
| <figure class="hero-visual" aria-hidden="true"> | |
| <canvas class="morph-canvas" id="morph"></canvas> | |
| <figcaption class="morph-labels"><span class="ml is-on" data-shape="0">Crystal lattice</span><span class="ml" data-shape="1">DNA double helix</span></figcaption> | |
| </figure> | |
| </div> | |
| </section> | |
| <section class="team-strip" aria-label="Who we are"> | |
| <div class="container"> | |
| <p class="team-title">Our team includes researchers from</p> | |
| <ul class="team-list"> | |
| <li>University of Maryland</li> | |
| <li>University of Georgia</li> | |
| <li>National University of Singapore</li> | |
| <li>Indiana University</li> | |
| <li>Carnegie Mellon University</li> | |
| <li>University of Pennsylvania</li> | |
| </ul> | |
| <p class="team-more">and we're looking for more collaborators</p> | |
| </div> | |
| </section> | |
| <section class="section" id="why" aria-labelledby="why-title"> | |
| <div class="container why-grid"> | |
| <div class="why-copy reveal"> | |
| <p class="kicker">Why we need you</p> | |
| <h2 id="why-title">Frontier AI agents still struggle with real science.</h2> | |
| <p class="sub">We built a test from 100 tasks based on real scientific software and ran eight frontier models on it. The best one solved one task in five. Life-science tasks were solved once in 120 attempts.</p> | |
| <p class="sub">Most training data for AI agents comes from software engineering, so the workflows scientists actually run are barely covered. Closing that gap takes people who know what a correct result looks like in their field.</p> | |
| </div> | |
| <div class="stat-stack reveal"> | |
| <div class="stat"><strong>20%</strong><span>Success rate of the best of eight frontier models on our held-out scientific-software test</span></div> | |
| <div class="stat stat-warm"><strong>1 in 120</strong><span>Attempts that succeeded on life & molecular science tasks, across all eight models</span></div> | |
| <div class="stat stat-cool"><strong>98%</strong><span>Agreement between our automated checks and blinded human reviewers</span></div> | |
| </div> | |
| </div> | |
| </section> | |
| <section class="section section-tint" id="who" data-nav="why" aria-labelledby="who-title"> | |
| <div class="container"> | |
| <div class="section-head reveal"> | |
| <p class="kicker">Who we're looking for</p> | |
| <h2 id="who-title">Graduate students in biology and materials science.</h2> | |
| <p class="sub">Master's or PhD, wet lab or computational, any level of AI experience. If you script, simulate, image, or analyze data as part of your research, you have exactly the knowledge we're missing.</p> | |
| </div> | |
| <div class="field-grid"> | |
| <article class="field-card field-bio reveal"> | |
| <div class="field-icon" aria-hidden="true"> | |
| <svg viewBox="0 0 48 48"><path d="M16 6c0 12 16 12 16 24s-16 12-16 12M32 6c0 12-16 12-16 24s16 12 16 12" fill="none" stroke="currentColor" stroke-width="2.6" stroke-linecap="round"/><path d="M18 12h12M19.5 19h9M19.5 29h9M18 36h12" stroke="currentColor" stroke-width="2.2" stroke-linecap="round" opacity=".55"/></svg> | |
| </div> | |
| <h3>Biology</h3> | |
| <p>Cell and molecular biology, genomics and bioinformatics, structural biology, neuroscience, microbiology, and ecology.</p> | |
| <p class="tools-label">Tools you might use</p> | |
| <ul class="tool-chips"> | |
| <li>Fiji / ImageJ</li><li>CellProfiler</li><li>Biopython</li><li>BLAST</li><li>Snakemake</li><li>R / Bioconductor</li><li>AlphaFold</li><li>PyMOL</li><li>GROMACS</li> | |
| </ul> | |
| </article> | |
| <article class="field-card field-mat reveal"> | |
| <div class="field-icon" aria-hidden="true"> | |
| <svg viewBox="0 0 48 48"><path d="M10 16l14-8 14 8v16l-14 8-14-8z" fill="none" stroke="currentColor" stroke-width="2.6" stroke-linejoin="round"/><path d="M10 16l14 8 14-8M24 24v16" fill="none" stroke="currentColor" stroke-width="2.2" stroke-linejoin="round" opacity=".55"/><g fill="currentColor"><circle cx="24" cy="8" r="3"/><circle cx="10" cy="16" r="3"/><circle cx="38" cy="16" r="3"/><circle cx="24" cy="24" r="3"/><circle cx="10" cy="32" r="3"/><circle cx="38" cy="32" r="3"/><circle cx="24" cy="40" r="3"/></g></svg> | |
| </div> | |
| <h3>Materials science</h3> | |
| <p>Computational materials, characterization, polymers and soft matter, energy materials, and metals and ceramics.</p> | |
| <p class="tools-label">Tools you might use</p> | |
| <ul class="tool-chips"> | |
| <li>VASP</li><li>Quantum ESPRESSO</li><li>LAMMPS</li><li>ASE</li><li>pymatgen</li><li>Materials Project</li><li>GSAS-II</li><li>OVITO</li><li>COMSOL</li> | |
| </ul> | |
| </article> | |
| </div> | |
| <p class="who-note reveal">Working in a neighboring field, like chemistry, biophysics, or chemical engineering? We'd still love to hear from you.</p> | |
| </div> | |
| </section> | |
| <section class="section" id="your-research" aria-labelledby="yr-title"> | |
| <div class="container"> | |
| <div class="section-head reveal"> | |
| <p class="kicker">Let's talk about your research</p> | |
| <h2 id="yr-title">Where does AI help you today, and where does it fall short?</h2> | |
| <p class="sub">We want to understand research from the inside: what fills your week, which steps AI already speeds up, and which ones it gets wrong. Here is our starting picture of a typical project. Help us correct it.</p> | |
| </div> | |
| <div class="explorer reveal" id="explorer"> | |
| <div class="explorer-top"> | |
| <div class="field-toggle" role="group" aria-label="Choose a field"> | |
| <button type="button" class="ft-btn" data-field="bio" aria-pressed="true">Biology</button> | |
| <button type="button" class="ft-btn" data-field="mat" aria-pressed="false">Materials science</button> | |
| </div> | |
| <p class="explorer-hint">Pick a stage of a typical project.</p> | |
| </div> | |
| <div class="explorer-panel" data-field="bio"> | |
| <h3 class="panel-title">Biology</h3> | |
| <article class="stage"> | |
| <h4 class="stage-title">Reading & planning</h4> | |
| <dl class="stage-grid"> | |
| <div class="sg sg-work"><dt>What it looks like</dt><dd>Keeping up with papers and preprints, comparing protocols, and planning the next experiment.</dd></div> | |
| <div class="sg sg-gap"><dt>Where AI falls short</dt><dd>Confident summaries with missing or invented citations, and little sense of which protocol details actually matter.</dd></div> | |
| <div class="sg sg-build"><dt>What we'd build together</dt><dd>Tasks that check an agent's answers against the papers and protocols you rely on.</dd></div> | |
| </dl> | |
| </article> | |
| <article class="stage"> | |
| <h4 class="stage-title">Designing experiments</h4> | |
| <dl class="stage-grid"> | |
| <div class="sg sg-work"><dt>What it looks like</dt><dd>Designing primers, guide RNAs, constructs, and controls, often across several separate tools.</dd></div> | |
| <div class="sg sg-gap"><dt>Where AI falls short</dt><dd>Chaining specialized tools together and respecting organism- and lab-specific constraints.</dd></div> | |
| <div class="sg sg-build"><dt>What we'd build together</dt><dd>Design tasks scored by the same tools and rules you use to sanity-check a design.</dd></div> | |
| </dl> | |
| </article> | |
| <article class="stage"> | |
| <h4 class="stage-title">Processing data</h4> | |
| <dl class="stage-grid"> | |
| <div class="sg sg-work"><dt>What it looks like</dt><dd>Sequencing pipelines, microscopy segmentation, and flow-cytometry or plate-reader data.</dd></div> | |
| <div class="sg sg-gap"><dt>Where AI falls short</dt><dd>Installing tools, wiring up multi-step pipelines, and catching silent errors in the output.</dd></div> | |
| <div class="sg sg-build"><dt>What we'd build together</dt><dd>Pipelines rebuilt from your real runs, with checks on the numbers, not just on whether files exist.</dd></div> | |
| </dl> | |
| </article> | |
| <article class="stage"> | |
| <h4 class="stage-title">Analysis & modeling</h4> | |
| <dl class="stage-grid"> | |
| <div class="sg sg-work"><dt>What it looks like</dt><dd>Statistics, structure prediction, docking, and molecular dynamics.</dd></div> | |
| <div class="sg sg-gap"><dt>Where AI falls short</dt><dd>Choosing sensible methods and parameters, and noticing when a result is biologically implausible.</dd></div> | |
| <div class="sg sg-build"><dt>What we'd build together</dt><dd>Analyses checked on hidden test cases, so an agent has to get the method right, not just the format.</dd></div> | |
| </dl> | |
| </article> | |
| <article class="stage"> | |
| <h4 class="stage-title">Figures & writing</h4> | |
| <dl class="stage-grid"> | |
| <div class="sg sg-work"><dt>What it looks like</dt><dd>Figures, methods sections, and notebooks that others can rerun.</dd></div> | |
| <div class="sg sg-gap"><dt>Where AI falls short</dt><dd>Methods text that drifts from what the code actually did.</dd></div> | |
| <div class="sg sg-build"><dt>What we'd build together</dt><dd>Tasks where every figure and methods detail has to match the analysis that produced it.</dd></div> | |
| </dl> | |
| </article> | |
| </div> | |
| <div class="explorer-panel" data-field="mat"> | |
| <h3 class="panel-title">Materials science</h3> | |
| <article class="stage"> | |
| <h4 class="stage-title">Reading & planning</h4> | |
| <dl class="stage-grid"> | |
| <div class="sg sg-work"><dt>What it looks like</dt><dd>Finding synthesis routes, processing conditions, and property data scattered across many papers.</dd></div> | |
| <div class="sg sg-gap"><dt>Where AI falls short</dt><dd>Pulling values out of tables and figures without mixing up units, samples, or conditions.</dd></div> | |
| <div class="sg sg-build"><dt>What we'd build together</dt><dd>Extraction tasks checked against values curated by people who know the field.</dd></div> | |
| </dl> | |
| </article> | |
| <article class="stage"> | |
| <h4 class="stage-title">Setting up simulations</h4> | |
| <dl class="stage-grid"> | |
| <div class="sg sg-work"><dt>What it looks like</dt><dd>Building DFT or molecular-dynamics inputs, running convergence tests, and writing cluster job scripts.</dd></div> | |
| <div class="sg sg-gap"><dt>Where AI falls short</dt><dd>Physically sensible parameters, pseudopotential and force-field choices, and the quirks of each cluster.</dd></div> | |
| <div class="sg sg-build"><dt>What we'd build together</dt><dd>Simulation tasks checked by rerunning the real codes on inputs the agent never saw.</dd></div> | |
| </dl> | |
| </article> | |
| <article class="stage"> | |
| <h4 class="stage-title">Characterization</h4> | |
| <dl class="stage-grid"> | |
| <div class="sg sg-work"><dt>What it looks like</dt><dd>Processing XRD patterns, electron-microscopy images, and spectra.</dd></div> | |
| <div class="sg sg-gap"><dt>Where AI falls short</dt><dd>Peak fits and refinements that look reasonable but aren't physically right.</dd></div> | |
| <div class="sg sg-build"><dt>What we'd build together</dt><dd>Characterization pipelines rebuilt from real instrument data, with checks that respect measurement tolerances.</dd></div> | |
| </dl> | |
| </article> | |
| <article class="stage"> | |
| <h4 class="stage-title">Screening & data</h4> | |
| <dl class="stage-grid"> | |
| <div class="sg sg-work"><dt>What it looks like</dt><dd>Querying materials databases, featurizing structures, and screening candidates.</dd></div> | |
| <div class="sg sg-gap"><dt>Where AI falls short</dt><dd>Keeping structures, compositions, and units consistent from one tool to the next.</dd></div> | |
| <div class="sg sg-build"><dt>What we'd build together</dt><dd>Screening workflows checked against reference results from the codes you trust.</dd></div> | |
| </dl> | |
| </article> | |
| <article class="stage"> | |
| <h4 class="stage-title">Figures & writing</h4> | |
| <dl class="stage-grid"> | |
| <div class="sg sg-work"><dt>What it looks like</dt><dd>Phase diagrams, band structures, and methods that others can reproduce.</dd></div> | |
| <div class="sg sg-gap"><dt>Where AI falls short</dt><dd>Plots and text that drift from the underlying calculations.</dd></div> | |
| <div class="sg sg-build"><dt>What we'd build together</dt><dd>Tasks where every figure traces back to a verified calculation.</dd></div> | |
| </dl> | |
| </article> | |
| </div> | |
| </div> | |
| <div class="questions reveal"> | |
| <div class="questions-head"> | |
| <h3>Questions we'd love to ask you</h3> | |
| <p>No preparation needed. These are the kinds of things we'd talk about in a first conversation.</p> | |
| </div> | |
| <ol class="question-list"> | |
| <li>Which parts of your work do you think AI could help with?</li> | |
| <li>Where do you already use AI, and where have you stopped trusting it?</li> | |
| <li>What software, instruments, or pipelines does your work depend on?</li> | |
| <li>How would you evaluate whether an agent is doing the job well or badly?</li> | |
| <li>Which task would you hand to a reliable assistant tomorrow?</li> | |
| <li>What would an AI agent need to show before you'd publish its result?</li> | |
| </ol> | |
| </div> | |
| </div> | |
| </section> | |
| <section class="section section-tint" id="examples" aria-labelledby="ex-title"> | |
| <div class="container"> | |
| <div class="section-head reveal"> | |
| <p class="kicker">Demos from our previous work</p> | |
| <h2 id="ex-title">Agents working with real scientific software.</h2> | |
| <p class="sub">Each demo is a task from our research. An agent sees a few real runs of a scientific tool and writes a program that reproduces it. We then run that program on settings it has never seen and check every output against the real software.</p> | |
| </div> | |
| <div class="ex-grid" id="ex-grid"> | |
| <!--DEMO_CARDS:START--> | |
| <article class="ex-card reveal" data-index="0" data-kind="micro" tabindex="0" role="button" aria-label="Open demo: Microscopy cell segmentation and counting" style="--dc:#16a34a"> | |
| <div class="ex-thumb"><img src="assets/examples/micro-thumb.webp" alt="Microscopy label masks from public runs" width="740" height="394" loading="lazy"></div> | |
| <div class="ex-card-body"> | |
| <p class="ex-card-domain">Life & Molecular Sciences</p> | |
| <h3>Microscopy cell segmentation and counting</h3> | |
| <p>The same steps as counting nuclei or measuring cell shape in your own fluorescence images.</p> | |
| <ul class="ex-card-tracks"><li class="trk trk-rel">Biology</li><li class="trk trk-rel">Microscopy</li></ul> | |
| <ul class="ex-sw"><li>CellProfiler</li><li>scikit-image</li><li>Bio-Formats</li></ul> | |
| </div> | |
| <span class="ex-open" aria-hidden="true">Explore ↗</span> | |
| </article> | |
| <article class="ex-card reveal" data-index="1" data-kind="foam" tabindex="0" role="button" aria-label="Open demo: OpenFOAM 2-D dam-break CFD" style="--dc:#0891b2"> | |
| <div class="ex-thumb"><img src="assets/examples/foam-thumb.webp" alt="OpenFOAM dam-break mesh drawn from the public polyMesh" width="740" height="394" loading="lazy"></div> | |
| <div class="ex-card-body"> | |
| <p class="ex-card-domain">Physical & Engineering Simulation</p> | |
| <h3>OpenFOAM 2-D dam-break CFD</h3> | |
| <p>Meshing, boundary conditions, and solver settings: the setup behind any flow or transport simulation.</p> | |
| <ul class="ex-card-tracks"><li class="trk trk-rel">Simulation</li><li class="trk trk-rel">Fluid dynamics</li></ul> | |
| <ul class="ex-sw"><li>OpenFOAM</li></ul> | |
| </div> | |
| <span class="ex-open" aria-hidden="true">Explore ↗</span> | |
| </article> | |
| <article class="ex-card reveal" data-index="2" data-kind="stl" tabindex="0" role="button" aria-label="Open demo: OpenSCAD / BOSL2 parametric fan duct" style="--dc:#7c3aed"> | |
| <div class="ex-thumb"><img src="assets/examples/stl-thumb.webp" alt="Rendered fan-duct mesh from a public OpenSCAD run" width="740" height="394" loading="lazy"></div> | |
| <div class="ex-card-body"> | |
| <p class="ex-card-domain">Spatial, Visual & 3D</p> | |
| <h3>OpenSCAD / BOSL2 parametric fan duct</h3> | |
| <p>Parametric CAD for 3D-printed sample holders, fixtures, and flow cells.</p> | |
| <ul class="ex-card-tracks"><li class="trk trk-rel">Lab hardware</li><li class="trk trk-rel">CAD</li></ul> | |
| <ul class="ex-sw"><li>OpenSCAD</li><li>BOSL2</li></ul> | |
| </div> | |
| <span class="ex-open" aria-hidden="true">Explore ↗</span> | |
| </article> | |
| <article class="ex-card reveal" data-index="3" data-kind="astro" tabindex="0" role="button" aria-label="Open demo: Astronomy FITS/WCS cutout and photometry" style="--dc:#db2777"> | |
| <div class="ex-thumb"><img src="assets/examples/astro-thumb.webp" alt="Astronomy FITS cutouts from public runs" width="740" height="394" loading="lazy"></div> | |
| <div class="ex-card-body"> | |
| <p class="ex-card-domain">Earth & Space Sciences</p> | |
| <h3>Astronomy FITS/WCS cutout and photometry</h3> | |
| <p>Aperture photometry and background subtraction, the same math as quantifying spots or bands in noisy images.</p> | |
| <ul class="ex-card-tracks"><li class="trk trk-rel">Imaging</li><li class="trk trk-rel">Signal extraction</li></ul> | |
| <ul class="ex-sw"><li>Astropy</li><li>Photutils</li></ul> | |
| </div> | |
| <span class="ex-open" aria-hidden="true">Explore ↗</span> | |
| </article> | |
| <article class="ex-card reveal" data-index="4" data-kind="color" tabindex="0" role="button" aria-label="Open demo: Color-space and tone-curve pipeline" style="--dc:#7c3aed"> | |
| <div class="ex-thumb"><img src="assets/examples/color-thumb.webp" alt="Color-pipeline previews from public runs" width="740" height="394" loading="lazy"></div> | |
| <div class="ex-card-body"> | |
| <p class="ex-card-domain">Spatial, Visual & 3D</p> | |
| <h3>Color-space and tone-curve pipeline</h3> | |
| <p>The color-space math behind calibrated imaging, colorimetric assays, and optical characterization.</p> | |
| <ul class="ex-card-tracks"><li class="trk trk-rel">Imaging</li><li class="trk trk-rel">Optics</li></ul> | |
| <ul class="ex-sw"><li>colour-science</li><li>ImageMagick</li><li>scikit-image</li></ul> | |
| </div> | |
| <span class="ex-open" aria-hidden="true">Explore ↗</span> | |
| </article> | |
| <article class="ex-card reveal" data-index="5" data-kind="ocr" tabindex="0" role="button" aria-label="Open demo: Tesseract table-image OCR" style="--dc:#d97706"> | |
| <div class="ex-thumb"><img src="assets/examples/ocr-thumb.webp" alt="OCR page preview from a public run" width="740" height="394" loading="lazy"></div> | |
| <div class="ex-card-body"> | |
| <p class="ex-card-domain">Data & Document Workflows</p> | |
| <h3>Tesseract table-image OCR</h3> | |
| <p>Reading numbers out of scanned tables, the first step in mining property data from papers.</p> | |
| <ul class="ex-card-tracks"><li class="trk trk-rel">Literature</li><li class="trk trk-rel">Data extraction</li></ul> | |
| <ul class="ex-sw"><li>Pandoc</li><li>Poppler</li><li>Tesseract</li><li>QPDF</li></ul> | |
| </div> | |
| <span class="ex-open" aria-hidden="true">Explore ↗</span> | |
| </article> | |
| <article class="ex-card reveal is-extra" data-index="6" data-kind="robot" tabindex="0" role="button" aria-label="Open demo: MuJoCo / URDF mass-inertia repair" style="--dc:#0891b2"> | |
| <div class="ex-thumb"><img src="assets/examples/robot-thumb.webp" alt="MuJoCo joint-position trajectories across public scenarios" width="740" height="394" loading="lazy"></div> | |
| <div class="ex-card-body"> | |
| <p class="ex-card-domain">Physical & Engineering Simulation</p> | |
| <h3>MuJoCo / URDF mass-inertia repair</h3> | |
| <p>Physical models with masses, inertias, and joint limits, as used to simulate lab robots before running them.</p> | |
| <ul class="ex-card-tracks"><li class="trk trk-rel">Lab automation</li><li class="trk trk-rel">Physics</li></ul> | |
| <ul class="ex-sw"><li>MuJoCo</li><li>URDF/MJCF tooling</li></ul> | |
| </div> | |
| <span class="ex-open" aria-hidden="true">Explore ↗</span> | |
| </article> | |
| <article class="ex-card reveal is-extra" data-index="7" data-kind="blender" tabindex="0" role="button" aria-label="Open demo: Blender cloth flag scene" style="--dc:#7c3aed"> | |
| <div class="ex-thumb"><img src="assets/examples/blender-thumb.webp" alt="Blender cloth scene rotation dial across public scenarios" width="740" height="394" loading="lazy"></div> | |
| <div class="ex-card-body"> | |
| <p class="ex-card-domain">Spatial, Visual & 3D</p> | |
| <h3>Blender cloth flag scene</h3> | |
| <p>A physics simulation whose outcome shifts with wind, constraints, and resolution.</p> | |
| <ul class="ex-card-tracks"><li class="trk trk-rel">Simulation</li><li class="trk trk-rel">Physics</li></ul> | |
| <ul class="ex-sw"><li>Blender</li></ul> | |
| </div> | |
| <span class="ex-open" aria-hidden="true">Explore ↗</span> | |
| </article> | |
| <article class="ex-card reveal is-extra" data-index="8" data-kind="drc" tabindex="0" role="button" aria-label="Open demo: KiCad LED-matrix driver PCB" style="--dc:#7c3aed"> | |
| <div class="ex-thumb"><img src="assets/examples/drc-thumb.webp" alt="KiCad board with design-rule findings plotted at their positions" width="740" height="394" loading="lazy"></div> | |
| <div class="ex-card-body"> | |
| <p class="ex-card-domain">Spatial, Visual & 3D</p> | |
| <h3>KiCad LED-matrix driver PCB</h3> | |
| <p>Design-rule checks on circuit boards, like the custom electronics inside lab instruments.</p> | |
| <ul class="ex-card-tracks"><li class="trk trk-rel">Instrumentation</li><li class="trk trk-rel">Electronics</li></ul> | |
| <ul class="ex-sw"><li>KiCad</li></ul> | |
| </div> | |
| <span class="ex-open" aria-hidden="true">Explore ↗</span> | |
| </article> | |
| <article class="ex-card reveal is-extra" data-index="9" data-kind="container" tabindex="0" role="button" aria-label="Open demo: Container SBOM and supply-chain attestation" style="--dc:#2563eb"> | |
| <div class="ex-thumb"><img src="assets/examples/container-thumb.webp" alt="Container image layers and attestation from a public run" width="740" height="394" loading="lazy"></div> | |
| <div class="ex-card-body"> | |
| <p class="ex-card-domain">Software & Formal Systems</p> | |
| <h3>Container SBOM and supply-chain attestation</h3> | |
| <p>Recording exactly which software went into an environment, the backbone of reproducible analysis.</p> | |
| <ul class="ex-card-tracks"><li class="trk trk-rel">Reproducibility</li><li class="trk trk-rel">Software</li></ul> | |
| <ul class="ex-sw"><li>Hadolint</li><li>Syft</li><li>Skopeo</li></ul> | |
| </div> | |
| <span class="ex-open" aria-hidden="true">Explore ↗</span> | |
| </article> | |
| <!--DEMO_CARDS:END--> | |
| </div> | |
| <div class="ex-more"> | |
| <button class="btn btn-ghost" id="ex-more" type="button" aria-expanded="false">Show more demos</button> | |
| </div> | |
| </div> | |
| </section> | |
| <section class="section" id="collaborate" aria-labelledby="collab-title"> | |
| <div class="container"> | |
| <div class="section-head reveal"> | |
| <p class="kicker">How we'd work together</p> | |
| <h2 id="collab-title">From one conversation to a shared paper.</h2> | |
| <p class="sub">Start small and go as far as you like. Every step helps, and you decide how involved you want to be.</p> | |
| </div> | |
| <ol class="steps"> | |
| <li class="step reveal"><span class="step-num">1</span><h3>Talk</h3><p>An informal conversation about your research, your tools, and where AI helps or gets in the way.</p></li> | |
| <li class="step reveal"><span class="step-num">2</span><h3>Pick a workflow</h3><p>Together we choose a real piece of your work: an analysis script, a simulation setup, or a data pipeline.</p></li> | |
| <li class="step reveal"><span class="step-num">3</span><h3>Turn it into a test</h3><p>We package it as a task an agent can attempt, with checks you agree are scientifically right.</p></li> | |
| <li class="step reveal"><span class="step-num">4</span><h3>Improve and publish</h3><p>We measure today's agents, train better ones, and write up what we learn together.</p></li> | |
| </ol> | |
| <div class="collab-grid"> | |
| <div class="give-card reveal"> | |
| <h3>What you'd bring</h3> | |
| <ul class="check-list"> | |
| <li>How research in your field actually gets done</li> | |
| <li>Real workflows: scripts, simulations, and pipelines</li> | |
| <li>Judgment about what counts as a correct result</li> | |
| <li>Feedback on what agents get right and wrong</li> | |
| </ul> | |
| </div> | |
| <div class="give-card give-card-accent reveal"> | |
| <h3>What you'd get</h3> | |
| <ul class="check-list"> | |
| <li>Co-authorship on the resulting paper for substantial contributions</li> | |
| <li>A first look at how today's AI agents handle your own workflows</li> | |
| <li>A packaged, tested version of the workflow you contributed</li> | |
| <li>Hands-on experience with how AI agents are trained and evaluated</li> | |
| </ul> | |
| </div> | |
| </div> | |
| <div class="ways reveal"> | |
| <h3>Ways to take part</h3> | |
| <div class="ways-grid"> | |
| <div class="way"><span class="way-tag">One conversation</span><p>Tell us how you work and where AI lets you down.</p></div> | |
| <div class="way"><span class="way-tag">Contribute a workflow</span><p>Help turn a piece of your research into a task with real checks.</p></div> | |
| <div class="way"><span class="way-tag">Co-author</span><p>Join the analysis and the writing of the paper.</p></div> | |
| </div> | |
| </div> | |
| </div> | |
| </section> | |
| <section class="section section-tint" id="papers" aria-labelledby="papers-title"> | |
| <div class="container"> | |
| <div class="section-head reveal"> | |
| <p class="kicker">Our research so far</p> | |
| <h2 id="papers-title">Results from our recent papers.</h2> | |
| <p class="sub">The demos above come from this work. Each paper tackles one part of the problem: building tasks that check results reliably, making tasks harder over time, and turning an agent's successes into better training data.</p> | |
| </div> | |
| <div class="paper-list"> | |
| <article class="paper reveal"> | |
| <div class="paper-main"> | |
| <p class="paper-meta"><span class="paper-tag">Scientific software</span>arXiv · October 2026</p> | |
| <h3><a href="https://arxiv.org/abs/2610.02710" target="_blank" rel="noopener">Self-Supervised Scaling of Terminal Environments for Scientific Domains</a></h3> | |
| <p class="paper-authors">Zhongzhi Li, Yucheng Shi, Zongxia Li, et al.</p> | |
| <p class="paper-summary">Builds agent tasks directly from existing scientific software. The agent sees a few real runs and must reproduce the workflow, and hidden runs check the result. The tasks span six domains, including life and molecular sciences, physical and engineering simulation, and earth and space sciences.</p> | |
| <p class="paper-you"><strong>Why it matters to you:</strong> this is how we'd turn your workflows into tasks.</p> | |
| <div class="paper-links"><a class="btn btn-ghost btn-sm" href="https://arxiv.org/abs/2610.02710" target="_blank" rel="noopener">Read the paper</a></div> | |
| </div> | |
| <div class="paper-results"> | |
| <h4>Key results</h4> | |
| <div class="res"> | |
| <p class="res-label">Fine-tuning on these tasks, Terminal-Bench 2</p> | |
| <div class="rb rb-base" style="--v:47.9"><i></i><span>Before 47.9%</span></div> | |
| <div class="rb rb-ours" style="--v:53.6"><i></i><span>After 53.6%</span></div> | |
| </div> | |
| <div class="res"> | |
| <p class="res-label">Long-horizon terminal tasks</p> | |
| <div class="rb rb-base" style="--v:20.7"><i></i><span>Before 20.7%</span></div> | |
| <div class="rb rb-ours" style="--v:27.7"><i></i><span>After 27.7%</span></div> | |
| </div> | |
| <div class="res"> | |
| <p class="res-label">Automated checks agree with blinded human reviewers</p> | |
| <div class="rb rb-ours" style="--v:98"><i></i><span>98%</span></div> | |
| </div> | |
| </div> | |
| </article> | |
| <article class="paper reveal"> | |
| <div class="paper-main"> | |
| <p class="paper-meta"><span class="paper-tag">Harder tasks</span>arXiv · August 2026</p> | |
| <h3><a href="https://arxiv.org/abs/2608.05466" target="_blank" rel="noopener">Recursive Synthesis for Long-Horizon Terminal Tasks</a></h3> | |
| <p class="paper-authors">Zhongzhi Li, Yucheng Shi, Zongxia Li, et al.</p> | |
| <p class="paper-summary">Grows simple, verified tasks into long, multi-step ones, round after round. Every new task is checked in a fresh sandbox before it is kept, at roughly $0.05 per task.</p> | |
| <p class="paper-you"><strong>Why it matters to you:</strong> agents learn to finish whole projects, not just single commands.</p> | |
| <div class="paper-links"><a class="btn btn-ghost btn-sm" href="https://arxiv.org/abs/2608.05466" target="_blank" rel="noopener">Read the paper</a></div> | |
| </div> | |
| <div class="paper-results"> | |
| <h4>Key results</h4> | |
| <div class="res"> | |
| <p class="res-label">Same strong model, tasks from round 1 vs. round 15</p> | |
| <div class="rb rb-base" style="--v:90"><i></i><span>Round 1 solved 90%</span></div> | |
| <div class="rb rb-warm" style="--v:2.5"><i></i><span>Round 15 solved 2.5%</span></div> | |
| </div> | |
| <div class="res"> | |
| <p class="res-label">Median reference solution length</p> | |
| <div class="rb rb-base" style="--v:17.9"><i></i><span>67 lines</span></div> | |
| <div class="rb rb-ours" style="--v:100"><i></i><span>374 lines</span></div> | |
| </div> | |
| <div class="res"> | |
| <p class="res-label">Reinforcement learning on these tasks, Terminal-Bench Hard</p> | |
| <div class="rb rb-base" style="--v:22.7"><i></i><span>Before 22.7%</span></div> | |
| <div class="rb rb-ours" style="--v:32"><i></i><span>After 32.0% (+41% relative)</span></div> | |
| </div> | |
| </div> | |
| </article> | |
| <article class="paper reveal"> | |
| <div class="paper-main"> | |
| <p class="paper-meta"><span class="paper-tag">Better training data</span>arXiv · October 2026</p> | |
| <h3><a href="https://arxiv.org/abs/2610.02826" target="_blank" rel="noopener">Scaling Trajectories for Complex Tasks through Recursive Self-Rewrite</a></h3> | |
| <p class="paper-authors">Zongxia Li, Yucheng Shi, Zhongzhi Li, et al.</p> | |
| <p class="paper-summary">Collects an agent's successful solutions from several different setups, then rewrites them into one consistent style the model can learn from. The rewritten data teaches the model far more than the raw successes do.</p> | |
| <p class="paper-you"><strong>Why it matters to you:</strong> one good solution to your task can become many training examples.</p> | |
| <div class="paper-links"> | |
| <a class="btn btn-ghost btn-sm" href="https://arxiv.org/abs/2610.02826" target="_blank" rel="noopener">Read the paper</a> | |
| <a class="btn btn-ghost btn-sm" href="https://huggingface.co/IntelligenceLab/RSR-27B" target="_blank" rel="noopener">Model on Hugging Face</a> | |
| </div> | |
| </div> | |
| <div class="paper-results"> | |
| <h4>Key results</h4> | |
| <div class="res"> | |
| <p class="res-label">Terminal-Bench 2, solved within three tries</p> | |
| <div class="rb rb-base" style="--v:57"><i></i><span>Before 57.0%</span></div> | |
| <div class="rb rb-ours" style="--v:74.2"><i></i><span>After 74.2%</span></div> | |
| </div> | |
| <div class="res"> | |
| <p class="res-label">Terminal-Bench Hard, solved within three tries</p> | |
| <div class="rb rb-base" style="--v:39"><i></i><span>Before 39.0%</span></div> | |
| <div class="rb rb-ours" style="--v:63"><i></i><span>After 63.0%</span></div> | |
| </div> | |
| <div class="res"> | |
| <p class="res-label">Terminal-Bench 4, solved within three tries</p> | |
| <div class="rb rb-base" style="--v:1.5"><i></i><span>Before 1.5%</span></div> | |
| <div class="rb rb-ours" style="--v:9.1"><i></i><span>After 9.1%</span></div> | |
| </div> | |
| </div> | |
| </article> | |
| </div> | |
| </div> | |
| </section> | |
| <section class="section" id="faq" aria-labelledby="faq-title"> | |
| <div class="container faq-grid"> | |
| <div class="faq-head reveal"> | |
| <h2 class="kicker" id="faq-title">Questions</h2> | |
| <p class="sub">Anything else? Ask us directly. We're happy to talk it through.</p> | |
| </div> | |
| <div class="faq-list reveal"> | |
| <details open> | |
| <summary>Who are you?</summary> | |
| <p>AI researchers from the University of Maryland, the University of Georgia, the National University of Singapore, Indiana University, Carnegie Mellon University, and the University of Pennsylvania. We build and train AI agents that work with real software, we publish our results on arXiv, and we're looking for more collaborators.</p> | |
| </details> | |
| <details> | |
| <summary>Do I need machine-learning experience?</summary> | |
| <p>No. We bring the AI side. What we need is your knowledge of how the science is done and how you know a result is right.</p> | |
| </details> | |
| <details> | |
| <summary>How much time does it take?</summary> | |
| <p>It's up to you. A single conversation already helps us. Contributing a workflow takes more time, and we'll plan it around your schedule.</p> | |
| </details> | |
| <details> | |
| <summary>Do I have to share unpublished data?</summary> | |
| <p>No. Tasks can be built from open-source software with public or synthetic inputs. We'll agree together on anything that goes into a paper.</p> | |
| </details> | |
| <details> | |
| <summary>Can I be an author on the paper?</summary> | |
| <p>Yes. Collaborators who contribute workflows, tasks, or analysis can be co-authors on the resulting paper, following standard authorship norms.</p> | |
| </details> | |
| <details> | |
| <summary>Who can join?</summary> | |
| <p>Graduate students, Master's or PhD, in biology, materials science, and related fields. Not sure you fit? Write to us anyway.</p> | |
| </details> | |
| </div> | |
| </div> | |
| </section> | |
| <section class="section join" id="join" aria-labelledby="join-title"> | |
| <div class="container join-grid"> | |
| <div class="join-copy reveal"> | |
| <p class="kicker">Get involved</p> | |
| <h2 id="join-title">Tell us about your research.</h2> | |
| <p class="sub">A few lines is plenty, and we'll reply to set up a conversation. Sending opens your email app with your answers filled in, so nothing is stored on this site.</p> | |
| <p class="join-alt">Prefer plain email? Write to <a class="mail-link" data-contact data-show-address data-subject="AI for Science collaboration" href="mailto:zli12321@terpmail.umd.edu,zl22754@uga.edu">zli12321@terpmail.umd.edu and zl22754@uga.edu</a>.</p> | |
| </div> | |
| <form class="join-form reveal" id="join-form" novalidate> | |
| <div class="row2"> | |
| <label>Name<input name="name" autocomplete="name" required></label> | |
| <label>University & program<input name="program" placeholder="e.g. PhD, Materials Science"></label> | |
| </div> | |
| <fieldset class="pill-set"> | |
| <legend>Field</legend> | |
| <div class="pills"> | |
| <label><input type="radio" name="field" value="Biology" checked><span>Biology</span></label> | |
| <label><input type="radio" name="field" value="Materials science"><span>Materials science</span></label> | |
| <label><input type="radio" name="field" value="Other"><span>Something else</span></label> | |
| </div> | |
| </fieldset> | |
| <label>Your research in a sentence<input name="topic" placeholder="e.g. segmenting neurons in light-sheet images"></label> | |
| <label>How do you use AI in your research today?<textarea name="ai_use" rows="3" placeholder="Writing code, reading papers, structure prediction, nothing yet…"></textarea></label> | |
| <fieldset class="pill-set"> | |
| <legend>How would you like to take part?</legend> | |
| <div class="pills"> | |
| <label><input type="checkbox" name="involve" value="A conversation" checked><span>A conversation</span></label> | |
| <label><input type="checkbox" name="involve" value="Contribute a workflow"><span>Contribute a workflow</span></label> | |
| <label><input type="checkbox" name="involve" value="Co-author the paper"><span>Co-author the paper</span></label> | |
| </div> | |
| </fieldset> | |
| <input type="hidden" name="interest" value=""> | |
| <div class="form-actions"> | |
| <button class="btn btn-dark" type="submit">Send to the team | |
| <svg viewBox="0 0 20 20" aria-hidden="true"><path d="M4 10h11m-4-4 4 4-4 4" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round"/></svg> | |
| </button> | |
| <p class="form-note" id="form-note" aria-live="polite"></p> | |
| </div> | |
| </form> | |
| </div> | |
| </section> | |
| </main> | |
| <footer class="footer"> | |
| <div class="container footer-inner"> | |
| <a class="brand" href="#top" aria-label="Back to top"> | |
| <svg class="brand-mark" viewBox="0 0 32 32" aria-hidden="true"> | |
| <rect width="32" height="32" rx="8" fill="url(#brand-grad)"/> | |
| <path d="M7 23h5v-5h5v-5h5" fill="none" stroke="#fff" stroke-width="2.6" stroke-linecap="round" stroke-linejoin="round"/> | |
| <path d="M19.6 9.4l2.2 2.2 4.2-4.4" fill="none" stroke="#fff" stroke-width="2.6" stroke-linecap="round" stroke-linejoin="round"/> | |
| </svg> | |
| <span>IntelligenceLab <em>AI for Science</em></span> | |
| </a> | |
| <nav class="footer-links" aria-label="Footer"> | |
| <a href="#your-research">Your research</a> | |
| <a href="#examples">Demos</a> | |
| <a href="#papers">Papers</a> | |
| <a href="#join">Get involved</a> | |
| </nav> | |
| <p class="footer-note">© <span data-year>2026</span> IntelligenceLab. Software names belong to their respective projects.</p> | |
| </div> | |
| </footer> | |
| <div class="ex-modal" id="ex-modal" hidden> | |
| <div class="ex-backdrop" data-ex-close></div> | |
| <div class="ex-sheet" role="dialog" aria-modal="true" aria-labelledby="exm-title"> | |
| <header class="ex-head"> | |
| <div> | |
| <p class="ex-domain" id="exm-domain"></p> | |
| <h3 id="exm-title"></h3> | |
| <ul class="ex-tracks" id="exm-tracks"></ul> | |
| </div> | |
| <div class="ex-nav"> | |
| <button class="icon-btn" type="button" data-ex-prev aria-label="Previous demo">←</button> | |
| <button class="icon-btn" type="button" data-ex-next aria-label="Next demo">→</button> | |
| <button class="icon-btn" type="button" data-ex-close aria-label="Close">✕</button> | |
| </div> | |
| </header> | |
| <div class="ex-body"> | |
| <figure class="ex-figure"> | |
| <div class="ex-live-hero" id="exm-live" aria-hidden="true" hidden></div> | |
| <figcaption class="ex-evidence-label">Evidence from the task's public runs</figcaption> | |
| <img id="exm-img" alt="" width="1600" height="939"> | |
| </figure> | |
| <p class="ex-desc" id="exm-desc"></p> | |
| <p class="ex-why" id="exm-why"></p> | |
| <div class="ex-facts"> | |
| <div><h4>Software</h4><ul id="exm-software"></ul></div> | |
| <div><h4>What the agent sees</h4><ul id="exm-artifacts"></ul></div> | |
| <div><h4>What we vary</h4><ul id="exm-axes"></ul></div> | |
| <div><h4>What gets checked</h4><ul id="exm-outputs"></ul></div> | |
| </div> | |
| <p class="ex-how">The agent sees public runs of the real software and writes a program that reproduces it. That program is then run on hidden settings of these parameters and must match the software's outputs, field by field.</p> | |
| <a class="btn btn-dark" id="exm-cta" href="#join">Work with tools like these? Let's talk | |
| <svg viewBox="0 0 20 20" aria-hidden="true"><path d="M4 10h11m-4-4 4 4-4 4" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round"/></svg> | |
| </a> | |
| </div> | |
| </div> | |
| </div> | |
| <script type="application/json" id="examples-data">[{"kind": "micro", "title": "Microscopy cell segmentation and counting", "domain": "Life & Molecular Sciences", "color": "#16a34a", "description": "Reconstruct cell segmentation and measurement behavior across channel, illumination, threshold, object size, and measurement feature.", "why": "The same steps as counting nuclei or measuring cell shape in your own fluorescence images.", "tags": ["Biology", "Microscopy"], "software": ["CellProfiler", "scikit-image", "Bio-Formats"], "artifacts": ["label-mask image", "object-measurements CSV"], "axes": ["channel selection", "illumination correction", "measurement feature", "object size range", "threshold method"], "outputs": ["object count", "area sum", "eccentricity sum", "threshold", "minimum size"], "img": "assets/examples/micro-detail.webp", "alt": "Microscopy label masks from public runs"}, {"kind": "foam", "title": "OpenFOAM 2-D dam-break CFD", "domain": "Physical & Engineering Simulation", "color": "#0891b2", "description": "Reconstruct a 2-D dam-break CFD workflow across water height, mesh, gravity, and simulation time.", "why": "Meshing, boundary conditions, and solver settings: the setup behind any flow or transport simulation.", "tags": ["Simulation", "Fluid dynamics"], "software": ["OpenFOAM"], "artifacts": ["blockMesh log", "polyMesh", "initial field files", "case dictionaries"], "axes": ["water height", "mesh", "gravity", "time"], "outputs": ["mesh statistics", "field summaries", "residual/probe behavior"], "img": "assets/examples/foam-detail.webp", "alt": "OpenFOAM dam-break mesh drawn from the public polyMesh"}, {"kind": "stl", "title": "OpenSCAD / BOSL2 parametric fan duct", "domain": "Spatial, Visual & 3D", "color": "#7c3aed", "description": "Regenerate a swept fan-duct mount STL whose inlet size, outlet size, bend angle, and rib count change the geometry.", "why": "Parametric CAD for 3D-printed sample holders, fixtures, and flow cells.", "tags": ["Lab hardware", "CAD"], "software": ["OpenSCAD", "BOSL2"], "artifacts": ["STL mesh", "geometry metrics JSON"], "axes": ["inlet size", "outlet size", "bend angle", "rib count"], "outputs": ["bounding box", "volume", "surface area", "connected components", "sampled surface distance"], "img": "assets/examples/stl-detail.webp", "alt": "Rendered fan-duct mesh from a public OpenSCAD run"}, {"kind": "astro", "title": "Astronomy FITS/WCS cutout and photometry", "domain": "Earth & Space Sciences", "color": "#db2777", "description": "Reconstruct FITS/WCS cutout and photometry behavior under aperture, background, clipping, threshold, and coordinate-frame interventions.", "why": "Aperture photometry and background subtraction, the same math as quantifying spots or bands in noisy images.", "tags": ["Imaging", "Signal extraction"], "software": ["Astropy", "Photutils"], "artifacts": ["FITS image", "photometry catalog CSV"], "axes": ["aperture radius", "background model", "sigma clip", "source threshold", "WCS frame"], "outputs": ["flux", "coordinate frame", "median", "shape", "standard deviation"], "img": "assets/examples/astro-detail.webp", "alt": "Astronomy FITS cutouts from public runs"}, {"kind": "color", "title": "Color-space and tone-curve pipeline", "domain": "Spatial, Visual & 3D", "color": "#7c3aed", "description": "Infer a color pipeline across color space, white point, tone curve, demosaicing, and compression quality.", "why": "The color-space math behind calibrated imaging, colorimetric assays, and optical characterization.", "tags": ["Imaging", "Optics"], "software": ["colour-science", "ImageMagick", "scikit-image"], "artifacts": ["color measurements JSON", "PPM preview"], "axes": ["color space", "white point", "tone curve", "demosaic method", "compression quality"], "outputs": ["mean color", "image shape", "XYZ tristimulus values"], "img": "assets/examples/color-detail.webp", "alt": "Color-pipeline previews from public runs"}, {"kind": "ocr", "title": "Tesseract table-image OCR", "domain": "Data & Document Workflows", "color": "#d97706", "description": "Reconstruct how DPI, font, language, metadata, and page range affect OCR behavior.", "why": "Reading numbers out of scanned tables, the first step in mining property data from papers.", "tags": ["Literature", "Data extraction"], "software": ["Pandoc", "Poppler", "Tesseract", "QPDF"], "artifacts": ["OCR summary JSON", "page-preview image"], "axes": ["DPI", "font", "language", "metadata", "page range"], "outputs": ["ink", "line count", "page size", "text digest", "text length"], "img": "assets/examples/ocr-detail.webp", "alt": "OCR page preview from a public run"}, {"kind": "robot", "title": "MuJoCo / URDF mass-inertia repair", "domain": "Physical & Engineering Simulation", "color": "#0891b2", "description": "Infer robot state and trajectory behavior under actuator, collision, pose, joint-limit, and target interventions.", "why": "Physical models with masses, inertias, and joint limits, as used to simulate lab robots before running them.", "tags": ["Lab automation", "Physics"], "software": ["MuJoCo", "URDF/MJCF tooling"], "artifacts": ["URDF model", "trajectory CSV", "control summary"], "axes": ["actuator gain", "collision margin", "initial pose", "joint limit", "trajectory target"], "outputs": ["gain", "nq", "nv", "qpos", "qvel"], "img": "assets/examples/robot-detail.webp", "alt": "MuJoCo joint-position trajectories across public scenarios"}, {"kind": "blender", "title": "Blender cloth flag scene", "domain": "Spatial, Visual & 3D", "color": "#7c3aed", "description": "Reconstruct cloth-scene behavior under wind, pin-group, frame, and subdivision interventions.", "why": "A physics simulation whose outcome shifts with wind, constraints, and resolution.", "tags": ["Simulation", "Physics"], "software": ["Blender"], "artifacts": ["scene graph JSON"], "axes": ["wind strength", "pin group", "frame", "cloth subdiv"], "outputs": ["evaluated objects", "frame", "materials", "node groups", "object counts", "rotation"], "img": "assets/examples/blender-detail.webp", "alt": "Blender cloth scene rotation dial across public scenarios"}, {"kind": "drc", "title": "KiCad LED-matrix driver PCB", "domain": "Spatial, Visual & 3D", "color": "#7c3aed", "description": "Reconstruct LED-matrix driver board outputs (DRC report and project files) under row, column, driver, and routing interventions.", "why": "Design-rule checks on circuit boards, like the custom electronics inside lab instruments.", "tags": ["Instrumentation", "Electronics"], "software": ["KiCad"], "artifacts": ["DRC report", "project file list"], "axes": ["rows", "cols", "driver", "routing"], "outputs": ["DRC report summary", "project file inventory"], "img": "assets/examples/drc-detail.webp", "alt": "KiCad board with design-rule findings plotted at their positions"}, {"kind": "container", "title": "Container SBOM and supply-chain attestation", "domain": "Software & Formal Systems", "color": "#2563eb", "description": "Infer coupled container base-image, package-set, label, entrypoint, and SBOM-scope behavior.", "why": "Recording exactly which software went into an environment, the backbone of reproducible analysis.", "tags": ["Reproducibility", "Software"], "software": ["Hadolint", "Syft", "Skopeo"], "artifacts": ["Dockerfile", "in-toto attestation", "SPDX SBOM"], "axes": ["base image", "entrypoint", "labels", "package set", "SBOM scope"], "outputs": ["attestation", "base", "layer count", "merkle", "vuln budget"], "img": "assets/examples/container-detail.webp", "alt": "Container image layers and attestation from a public run"}]</script> | |
| <script src="app.js" defer></script> | |
| <script src="live.js" defer></script> | |
| </body> | |
| </html> | |