Work
Agents, retrieval, on-device inference and vision, and the products built around them. Every number comes from an eval I ran. Case studies where they exist, source where they don’t.
ANU
Apple Swift Challenge 2026 Winner
A voice-first iOS assistant whose brain runs on the phone: Gemma 4 E4B through LiteRT, on-device classifiers routing every message, and the cloud only as a sanitized, optional fallback.
Case studyDual-LLM Personalization System
Final project for Northeastern’s graduate Generative AI course, written up as a NeurIPS-format paper
A three-layer router that classifies query complexity and dispatches each request to the right model, cutting cascade rate from 30.8% to 14.7%.
Case studyAgent Certifier
Grades agent code it did not write, without running it
A FastAPI and Next.js service that statically analyzes an AI agent codebase across 60 tests in 6 weighted categories, from security to ethics, and returns a certification level with a report of what failed and why.
Case studyMore projects
RAG from Scratch
A twelve-stage RAG ladder, each technique measured on BEIR SciFact against a bge-base dense baseline (nDCG@10 0.759) and re-tested with paired significance tests. LLM query rewriting made retrieval worse, a strong reranker made it better (+0.044 on dense, +0.110 on BM25), and most of the rest made no detectable difference. The guide walks through each result and why.
12 stages, every result significance-tested
Agentic Second Brain
Gemini extraction agent that parses order confirmations into a structured SQLite database, then answers order-history queries and surfaces re-order links through a chat UI.
deployed on DigitalOcean
AIFORG
A Palantir-style AI platform for a whole company: role-gated agents grounded in its own documents, with 14-category PII tokenization and a full audit trail.
new tool integration under 2 hours
Agentic Research Assistant
Multi-agent research loop with orchestrator, research, and reasoning agents over dual-layer ChromaDB memory, long-term plus working, retaining context across sessions.
3 agents, ArXiv-grounded
Notova
Privacy-first voice notes on iOS and macOS from one shared Swift core, a parallel Kotlin app on Android, and a backend architected so it can never receive audio.
3 platforms, 0 bytes of audio server-side
Meeting Bot
Local-first macOS meeting notetaker. Captures whatever audio is already playing, transcribes on-device with Whisper, ships three front-ends over one engine.
3 front-ends, $0 running cost
TryOn
Embeddable virtual try-on: a 360 degree phone video becomes a 3D body mesh, a brand’s size chart drives the fit, and a 2D photo mode warps garments onto the shopper’s own photo. Web, Android and iOS SDKs.
video to 3D body mesh, size-accurate fit
AutoFill Pro
Chrome extension that fills job applications across 46 tested platforms from 44 locally stored profile fields, with an optional AI fallback for custom questions. Contact details are redacted on-device before anything is sent to the AI.
46 platforms, 1,300 test cases
SkillGap AI
Skills-gap analysis platform mapping user profiles against live job-market embeddings. Three PMI San Francisco awards as sole developer.
500+ concurrent users
Real-time Lane Detection
LaneNet encoder-decoder with an EfficientNet backbone, profiled in PyTorch until it ran in real time.
99% accuracy at 150 FPS
Vision-Language Grounding
Zero-shot CLIP grounding with cross-modal attention visualization, Stable Diffusion generation, and embedding-space analysis.
4 pipelines, fully zero-shot
Real-time Human Pose Estimation
Published at IEEE AISP 2024. A lightweight keypoint regression head on MediaPipe landmark priors, with LSTM activity recognition across 11 activities.
85% mAP, 0.78 IoU, 30+ FPS
Hackathons and smaller builds
Omni-Axon
Stanford x DeepMind hackathon, April 2026. A master agent classifies each request, then answers, runs a saved workflow, or assembles a team of sub-agents with Drive and Docs tools.
agent builder, reusable workflows
Agent-Guidelines
Guideline-driven multi-agent orchestrator with persistent memory across runs and a live execution dashboard, so agent runs can be watched and diagnosed while they happen.
live execution dashboard
BugHound
CodePath AI110 capstone. I extended CodePath’s BugHound starter agent with an LLM-output quality gate, import-risk scoring, and a 20-check evaluation harness.
coursework: quality gate + eval harness
Neural Studio
Infinite-canvas workspace pairing Research and Creator modes on a single surface, built against the Gemini API.
Devpost hackathon build
Hospital Management System
Full-stack clinical portal covering patients, appointments, and billing, with JWT authentication and role-scoped access over MongoDB.
jwt auth, role-scoped access
Avigis Biopharma
The official website for Avigis Biopharma Private Limited, built and deployed end to end for an external client.
shipped for a client
More experiments and coursework live on GitHub.
Experience, education, and the stack I work in
Email, LinkedIn, and resume