Agents that survive production.

Robinhood· Agentic AI Team

current · May 2026 – Aug 2026

Machine Learning Engineer (Agentic) · Menlo Park, CA

Agent reliability and evaluation — the infrastructure layer that lets Robinhood ship AI into a regulated financial domain. Kafka-based news and market-data pipelines, Braintrust-driven evals, and post-training workflows (SFT on Databricks) spanning closed-source (GPT-4.1) and open-source models.

Python · OpenAI Agent SDK · Braintrust · Kubernetes · Kafka · Databricks · SFT

VYNN AI

Jul 2025 – Dec 2025

Founder & Software Engineer (Agentic) · Durham, NC

Built and ran a full-stack agentic financial analysis platform solo: LangGraph multi-agent backend, FastAPI orchestration layer, React dashboard, and production infrastructure on Hetzner Cloud. The pilot served ~500 users with full equity research (DCF modeling, news intelligence, automated reports) in under 7 minutes end-to-end; the product is live again and has grown to 5,000+ users.

Python · LangGraph · FastAPI · React · TypeScript · MongoDB · Redis · Kafka · Docker · Hetzner

Duke University

Aug 2025 – Apr 2026

Graduate Teaching Assistant · Durham, NC

Designed and ran CS 590 (Software Development Studio), where graduate students build AI debugging agents inspired by AutoCodeRover and deploy full-stack applications. Mentored teams in CS 408 and CS 390 on software architecture, DevOps, and LLM-oriented programming, shipping production software for outside clients.

Python · Docker · CI/CD · Git/GitLab · Web Assembly

AutoCodeRover (acquired by Sonar)

Aug 2024 – May 2025

Research Engineer · Singapore

Built the JetBrains IDE plugin end-to-end for autonomous code repair: GumTree-based 3-way AST merge, embedded SonarLint analysis, and real-time SSE streaming with per-step developer feedback. On the backend, added LLM-as-a-Judge self-improvement that lifted SWE-bench Verified to 51.6%, state-of-the-art among open-source agents at the time. The core technology was acquired by Sonar.

Kotlin · Python · IntelliJ Platform SDK · GumTree · SonarLint · JGit · OkHttp

Binance· Web3 Wallet Team

Jul 2025 – Oct 2025

Software Engineer · Singapore

Built backend validation infrastructure for Binance's Boosters campaign — automated API regression suites in CI, load-tested services to ~500K concurrent transactions via JMeter, and instrumented monitoring to catch consistency failures before production.

Java · REST APIs · iOS · Android SDKs · Postman · JMeter · CI/CD

NUS Undergraduate Research

Jan 2024 – Jul 2025

AI Researcher · Singapore

Led research as first author on a multi-agent AI framework for medical evidence synthesis. Designed and built LUMINA, a four-agent LLM framework that automates citation screening for medical systematic reviews — achieving 98.2% sensitivity and 87.9% specificity across 15 SRMAs (~150K citations) with a 35× reduction in false negatives vs. prior state-of-the-art.

Python · LangChain · CI/CD · GPT-4o-mini · GPT-o3-mini

Give it a ticker; get the report an equity analyst spends a day writing, in about seven minutes.

A LangGraph supervisor orchestrates seven agents: fundamentals, news intelligence, DCF valuation, and a three-layer recommendation engine in which the LLM never invents a number. I built all of it, about 50K lines across backend, orchestration, and frontend, and ran a ~500-user pilot on production infrastructure. It's back online now, free for everyone, and past 5,000 users.

app.vynnai.com (opens in new tab)·Architecture deep-dive·GitHub (opens in new tab)·Blog post

Autonomous code repair, from research prototype to acquired product.

I built the JetBrains plugin end-to-end, including a GumTree three-way AST merge that reconciles concurrent edits Git fundamentally can't, and the Self-Fix Agent that helped lift SWE-bench Verified from 38.4% to 51.6%. The core technology was acquired by Sonar. Their Foundation Agent, built on it, later reached 79.2%, top of the leaderboard at the February 2026 announcement.

Architecture deep-dive·ISSTA '24 paper (opens in new tab)·GitHub (opens in new tab)·Blog post

An operating system for agents, with git as the memory substrate.

Branches are execution contexts, commits are checkpoints, git reset --hard is rollback, and worktrees give parallel workers real filesystem isolation. v0 is shipped: three demos with committed transcripts, full cost telemetry, and CI-asserted hermetic rollback. The design argument is in the blog; the repo is the implementation.

Architecture deep-dive·GitHub (opens in new tab)·Design thesis

A four-agent framework that screens citations for medical systematic reviews the way human reviewers do.

Classifier, PICOS-guided screener, LLM-as-a-judge reviewer, and a self-correcting improvement loop. Across 15 published reviews and roughly 150,000 citations it reached 98.2% mean sensitivity with 87.9% specificity, a 35× reduction in missed studies against prior published baselines, at under a cent per article.

Research deep-dive·GitHub (opens in new tab)

Graduate course covering production software engineering, Docker, CI/CD, API design, and server deployment — culminating in students building an AI debugging agent (inspired by AutoCodeRover) and a full-stack social media application.

Mentoring teams on architecture, testing, DevOps, and full-stack development to deliver production-ready software for outside clients.

Leading weekly labs covering AI agents, LLM-oriented programming, Docker, APIs, and system design.