I'm a Computer Science student at Boston University, class of 2028, and I build things people actually use. Right now I'm engineering payments and AI document-extraction infrastructure at an early-stage startup, and studying machine learning through Cornell Tech's Break Through Tech program.

• Currently engineering at Grove Tax, working on payments infra and AI document extraction
• AI/ML Fellow at Break Through Tech @ Cornell Tech, and backend at Hack4Impact BU
• Previously co-founded Recost, an API cost-intelligence platform
• Eagle Scout · based in Boston, MA
• Connect with me on LinkedIn

• Shipping payments and AI extraction infrastructure at Grove Tax
• Building out Sapling Learn, now in closed beta at BU, with a pitch at the Neurotech Frontiers Summit this fall
• Deepening my ML work through the Break Through Tech industry challenge project
• Interested in developer tooling, fintech infrastructure, and anything where correctness actually matters

An AI study platform that builds a live knowledge graph as you learn. Three tutoring modes (Socratic, Expository, Teach-back), adaptive quizzes driven by per-concept mastery data, syllabus-to-assignment extraction, and study rooms for comparing progress with classmates. Currently in closed beta at BU.

Scale: 166 API endpoints · 47-table Postgres schema · 18 LLM agents · 1,050+ tests

Key Features:

• Live knowledge graph that grows as you learn new concepts
• Adaptive quiz generation from per-concept mastery data
• Syllabus parsing with automatic assignment tracking and calendar integration
• Study rooms for comparing progress with classmates
• Application-layer AES-256-GCM encryption on sensitive columns, with numeric-preserving decryption for grades

Tech Stack: TypeScript Next.js React Python FastAPI Google Gemini PostgreSQL Supabase D3.js

Live: saplinglearn.com | GitHub: View Project | Devpost: View Submission

An API cost-intelligence platform for developers: analyze codebase API usage, estimate cost, detect inefficiencies, and generate optimization suggestions. Co-founded and incorporated as a Delaware C-corp with a 3-person team; launched on Product Hunt. Won the Stripe-sponsored Best Web API track, the second-largest sponsor track at UIUC's flagship hackathon (677 participants).

Key Features:

• Serverless REST API with 30+ endpoints on Cloudflare Workers (Hono, D1), Google OAuth, API key management, KV-based rate limiting
• VS Code extension using web-tree-sitter to parse JS/TS/Python and detect N+1 hotspots, batching and caching opportunities, and cross-file call origins
• Node.js and Python SDKs intercepting outbound HTTP at runtime to capture cost and latency telemetry across 8+ providers
• Interactive dashboard with D3.js visualizations for API spend, risk heatmaps, and optimization insights
• Endpoint-level risk assessment with cacheable pattern detection and estimated savings calculations

Tech Stack: TypeScript Cloudflare Workers Hono D1 KV React D3.js VS Code Extension API Tree-sitter

GitHub: View Project | Devpost: View Submission

A node-cluster visualization platform that collects, deduplicates, clusters, and summarizes breaking news from across the web to mitigate media bias and show the complete picture of an event.

Key Features:

• Multi-source news collection via RSS feeds and web scraping
• Smart deduplication using sentence embeddings and FAISS vector indexing
• AI-powered clustering and headline generation with OpenAI GPT-4o
• Interactive D3.js node visualization for exploring news clusters

Tech Stack: TypeScript React D3.js Python Flask OpenAI GPT-4o Supabase FAISS

GitHub: View Project | Devpost: View Submission

A real-time crowd monitoring system using AI-powered drones to detect dangerous crowd formations and predict crowd crush events before they occur, giving authorities critical minutes to respond.

Key Features:

• Multi-source WebRTC streaming from drones with real-time person detection overlays
• YOLOv8 detection with high-accuracy bounding boxes
• Spatial density analysis using Gaussian kernel algorithms with normalized coordinates
• Predictive alerts and interactive D3.js risk heatmaps with time-series analytics

Tech Stack: TypeScript Next.js React D3.js Python FastAPI YOLOv8 aiortc WebRTC OpenCV PyTorch Cloudflare

GitHub: View Project | Devpost: View Submission

• TaxForm Automation: Gemini-powered extraction for W-2 and 1099-NEC/INT/DIV forms, built on Cloudflare Workers with source highlighting, arithmetic validation, and an audit trail. React TypeScript Vite Cloudflare Workers
• ETL Pipeline: modular Python ETL with automated validation, quality checks, and production logging. Python Pandas
• Stock Portfolio Insights: full-stack portfolio tracker with real-time analytics and AI-generated risk assessment. React Flask Python

Also working with: Model Context Protocol (MCP), RAG + FAISS, model evaluation, Tree-sitter, Apache Airflow, WebRTC, Hono.

• Best Web API (Stripe Track) | HackIllinois '26 | Recost
• AI Tutor Track Winner | CivicHacks BU '26 | Sapling
• Best Design | PennApps XXVI | noogie
• Best Use of Cloudflare | HackHarvard '25 | Eyrie
• Eagle Scout · HSF Scholar · MLT Career Prep Fellow · SEO Edge · Dean's List