MOHAMED AZARUDEEN - NEW YORK, USA
MS COMPUTER SCIENCE - STONY BROOK UNIVERSITY
LOCAL-FIRST AI · AGENTS · RAG · END-TO-END SYSTEMS
I'm Mohamed Azarudeen, a software engineer who builds AI systems end to end: the pipeline, the model integration, the API, and the interface it all lives in.
I'm currently pursuing a Master of Science in Computer Science at Stony Brook University, where I also work as a Research & Student Assistant. Before that, I worked as a Software Engineer Intern at LuminaView in New York, and as a Software Engineer I at Mavdero Techservices in Chennai, India.
The through-line in my work is local-first AI: systems powered by lightweight local LLMs, RAG pipelines with cited sources, multi-agent orchestration, and interfaces that make all of it feel effortless.
Stony Brook University
LuminaView
Mavdero Techservices
Upload a company's annual report (10-K/10-Q) and it gets converted to Markdown, semantically chunked, embedded, and indexed for hybrid search; an LLM extracts structured KPIs (income statement, balance sheet, risk factors, growth drivers) into Postgres, while a RAG chatbot answers questions grounded in the ingested filings. FastAPI backend, server-rendered dashboard, containerized and deployed live to Azure Kubernetes Service.
VIEW ON GITHUB ↗
Turns any blog post into a narrated podcast episode: paste a URL and watch it get scraped, summarized by a local LLM, and converted to speech, live, end to end. Rebuilt from a single-file script into a proper two-tier system with FastAPI + SQLite backend, Next.js/TypeScript/Tailwind frontend with real persistence and live progress updates, as a system-design learning project.
VIEW ON GITHUB ↗
A locally-runnable RAG-powered support system with a FastAPI backend, SQLite ticketing, and a Streamlit UI for answers documentation questions with cited sources, in both text and voice.
VIEW ON GITHUB ↗
A local, privacy-first emotional wellness app powered by lightweight Ollama LLMs (phi3, mistral, llama3) and zero cloud APIs, zero internet required. Four specialized agents: Therapist, Closure, Routine Planner, Brutal Honesty — collaborate to offer structured, empathetic breakup recovery support, including OCR-based chat screenshot analysis and a 7-day self-care plan.
VIEW ON GITHUB ↗
A Streamlit app for exploring and analyzing GitHub repositories using natural language, combining an MCP GitHub server with local Ollama-powered reasoning for a fully local-first AI-driven repo analysis assistant.
VIEW ON GITHUB ↗NO CLOUD DEPENDENCY.
NO DATA LEAVING THE MACHINE.
AGENTS, RETRIEVAL & VOICE — WHERE THE USER LIVES.
Master of Science, Computer Science