pancData3
An open-source MATLAB & Python pipeline for diffusion-weighted MRI analysis in MRI-guided radiotherapy.
Developed at Memorial Sloan Kettering under Dr. Eric Aliotta, pancData3 is the analysis backbone for an ongoing study of diffusion-weighted imaging in pancreatic-cancer radiotherapy. It unifies IVIM and ADC diffusion model fitting, deep-learning denoising, dose–volume histogram analysis against diffusion-defined GTV subvolumes, and cross-validated outcome modeling into a single reproducible pipeline.
Diffusion Modeling
- IVIM & ADC fitting
- GTV subvolume segmentation
- NIfTI / DICOM ingestion
- Per-voxel parameter maps
Deep Learning
- DnCNN image denoising
- IVIMnet parameter inference
- Patient-level cross-validation
- Reproducible training configs
Outcomes
- Dose–volume histogram analysis
- Competing-risks survival models
- Elastic-net feature selection
- Treatment-response prediction
MATLAB
Python
NumPy
PyTorch
scikit-learn
NIfTI / DICOM
DnCNN
IVIMnet
View on GitHub
→
PrismTask
Built for every kind of mind — a productivity app that works with your brain, not against it.
A cross-platform task manager that blends AI-assisted planning with wellness-focused productivity. Designed around neurodiversity and burnout prevention as first-class concerns rather than afterthoughts — combining task management, habit tracking, and a work–life balance engine in a single adaptive interface.
Core
- Projects, tags, subtasks
- Habit streaks & analytics
- Voice input & accessibility
- Home-screen widgets + timers
Wellness
- Work–Life Balance Engine
- Mood & energy tracking
- Burnout detection
- ND-friendly focus modes
AI (Pro)
- Eisenhower / Pomodoro+ coaching
- NLP batch operations
- AI briefing & weekly planning
- Conversation-to-task extraction
Kotlin
Jetpack Compose
React
TypeScript
FastAPI
PostgreSQL
Firebase
Claude AI
Learn about the app
→
Gamemasters Assistant
Campaign-prep for solo dungeon masters — Lazy DM, collaborative design, and proactive roleplaying in one tool.
A comprehensive campaign-prep and tableside orchestration web app that blends award-winning tabletop methodologies into a unified, zero-friction workflow. Synthesizing Mike Shea's Lazy DM framework, Collaborative Campaign Design, and Proactive Roleplaying, it enables Game Masters to seamlessly map character ambitions, manage faction clocks, track secrets, and project redacted logs to players in real-time.
Worldbuilding
- Factions, pantheons, settlements
- Faction clocks & session zero
- Shared world facts compiler
- Safety guidelines (lines/veils)
Methodology
- 8-step Lazy DM session prep
- Secrets & proactive RP patterns
- PC-driven goal mapping
- Character ambition alignment
Run-Time & Sync
- Encounter & NPC generators
- Real-time 1.5s Firestore sync
- XP difficulty budget calculators
- JSON backup & raw export
Next.js 15
TypeScript
Tailwind
Firebase
Learn about the app
→
Anneal Ambiance
Procedural ambient soundscapes synthesized via physical simulations of thermodynamic relaxation.
A web-based generative synthesizer that creates endless, non-repeating acoustic environments in real-time. By translating principles of physical annealing and chaotic dynamics into audio parameters, it modulates resonance, reverb, and microtonal shifts to relieve cognitive tension and foster deep-focus states. Features an interactive visualizer mapping sound propagation in simulated spaces.
Generative Synthesis
- Procedural FM/subtractive synthesis
- Microtonal scale modulation
- Web Audio API audio graph
- Real-time binaural spatialization
Thermodynamic Modeling
- Simulated annealing heat schedules
- Chaotic double pendulum dynamics
- Dynamic friction & entropy decay
- Noise-field frequency mapping
Interface & Visuals
- Interactive 3D WebGL sound visualizer
- Glassmorphic, low-contrast UI
- Responsive mobile-first layout
- Custom ambient noise presets
TypeScript
React
Web Audio API
Three.js
IndexedDB
Vite
View Project Details
→
averyLoop
An autonomous agent pipeline that audits, implements, reviews, and merges code with minimal human intervention.
A self-directed software-engineering pipeline that chains specialized agents through an audit → implement → review → merge loop. Backed by retrieval-augmented generation over the target codebase, it surfaces work, drafts changes, critiques its own diffs, and shepherds them to merge — closing the loop on routine engineering tasks without constant supervision.
Pipeline
- Audit & backlog triage
- Automated implementation
- Self-review of diffs
- Gated auto-merge
Retrieval
- RAG over the codebase
- Context-aware change drafting
- Symbol & dependency lookup
- Grounded code generation
Orchestration
- Multi-agent task routing
- Human-in-the-loop checkpoints
- CI-aware merge gating
- Reproducible run logs
Python
RAG
Claude AI
Git
View on GitHub
→