Explore my work
My published projects and research, laid out on one workbench and grouped by area. Pick anything to read what it is, how it was tested, and where to go next. Every case study is also listed on the homepage.
Drag to move around · Ctrl or Cmd + scroll, or the + and − buttons, zoom · Esc closes a project
-
AI & machine learning
Clage
A from-scratch neuroevolution engine and local browser workbench for inspecting agents, experiments, and deterministic replay.
Poster frame from the Clage Studio demo video. Python, FastAPI, JavaScript, Canvas · pytest, Playwright, mypy, Ruff
429 Python tests + 4 JavaScript tests + 21 browser workflows in CI · OR/AND/XOR 5/5; sine 0/5 with minimal initialization
-
AI & machine learning
PyOpt
Sep 2026 · Qwen3-4B, LoRA, Unsloth
LoRA-tuned a 4B model to rewrite Python functions to run faster, scored by running the code. Published v1.1 run: correct 40/40 → 40/40, faster 29/40 → 36/40, median speedup 3.57× → 5.65×; out-of-distribution probe base 9/10, LoRA 8/10. Caveat: 13 of the 40 test prompts leaked from training, so this is not a clean held-out result. The split is fixed in v1.2; the model is not retrained yet.
-
Applications
Nytr
A nutrition and training app built around Penn State dining data. Deterministic recommendations, a Python backend, and a native iOS client.
Poster frame from the Nytr demo video. Python, FastAPI, PostgreSQL (Supabase, row-level security) · Swift, SwiftUI, HealthKit
1,135 backend tests (1,124 passing in CI with Postgres) · 251 iOS XCTest cases
-
Applications
Talks
Record on Apple Watch; transcribe and summarize on iPhone; save to Notion. Built around recoverable transfers and persistent jobs.
Swift 6, SwiftUI, watchOS, iOS · WatchConnectivity · SpeechAnalyzer · Foundation Models · Notion API
57 XCTest cases passing in CI · iOS and watchOS builds on macOS runners
CI tests the code on each side of the watch–phone boundary; end-to-end behavior needs a real iPhone and Apple Watch.
-
Applications · AI & machine learning
MedClarity
2026 · React, TypeScript, Express, Gemini API
Ask one question across several medical PDFs. One Gemini request returns an answer with page-level citations, and each citation is checked against the loaded documents. 66 Vitest unit tests passing in CI.
-
Systems & developer tools
Devvy
One local daemon coordinates Discord presence across coding tools while keeping prompts, code, and full paths private.
JavaScript (Node.js), TypeScript integrations · Discord IPC · GitHub Actions
26 test scripts, all run in CI · releases only after CI passes · macOS released
Windows support is on main but ships with the next release, v4.1.0.
-
Research
Extracting adverse drug reaction reports from Reddit
Undergraduate research · Penn State · 2026–present
In progress
With Dr. Md Sultan Al Nahian, I’m studying how language models can identify reported drug–reaction relationships in Reddit discussions. The work compares direct prompting, retrieval-augmented generation, and extraction with verification against human annotations.
-
Other experiments
ScholarAI
Jun 2026 · 5-person team · Python, pgvector
Scheme-discovery platform from the USAII Global AI Hackathon 2026. Our team was a finalist among 6,081+ participants. My part was the data engineering: normalization, embeddings, and pgvector ingestion for 1,008 schemes.