Software Engineer & Founder
Agentic Systems • Distributed Backends • AI Developer Tooling
M.S. Computer Science (AI/ML), Duke University

I am a Software Engineer and M.S. Computer Science (AI/ML) student at Duke University, focused on building production-grade agentic AI systems—from multi-agent orchestration and LLM tooling to the distributed backends that run them reliably at scale.

I am the sole founder and engineer of VYNN AI (https://vynnai.com), an agentic financial analyst platform built end-to-end, spanning multi-agent orchestration (LangGraph), real-time market data pipelines, and autonomous LLM-driven analysis deployed in production.

Previously, I designed core components of AutoCodeRover, an autonomous code repair system acquired by Sonar, integrating agentic reasoning directly into JetBrains IDEs and improving autonomous repair performance to 46% on SWE-bench Verified. I have also conducted research on multi-agent LLM frameworks for large-scale medical text mining, currently under review at NEJM AI.

I am interested in building autonomous AI agents that operate reliably in real-world systems, with an emphasis on correctness, observability, and production readiness over demo performance.

🔗 https://vynnai.com
Founder & Software Engineer

• Built and deployed an agentic financial analysis platform as sole engineer, orchestrating multi-agent workflows with LangGraph, tool-augmented reasoning, and Model Context Protocol (MCP) self-retrieval to generate structured long-form financial reports
• Architected production backends on Hetzner Cloud using FastAPI (async), Redis, MongoDB, Dockerized services, and Caddy, enabling real-time SSE/WebSocket streaming and fault-tolerant async job orchestration
• Designed a full valuation engine supporting financial projections, DCF modeling, sensitivity analysis, and automated HTML/PDF report generation via a React/TypeScript dashboard
• Optimized parallel agent execution with caching and scheduling, reducing end-to-end analysis latency by 72% in production

🔗 https://www.sonarsource.com/company/press-releases/sonar-acquires-autocoderover-to-supercharge-developers-with-ai-agents/
Software Engineer · Acquired by Sonar

• Built an end-to-end JetBrains IDE extension in Kotlin, integrating SonarLint, PSI inspections, REST APIs, and multithreaded background workers with UI-thread synchronization
• Designed an AST-level Patch Alignment mechanism using GumTree to reconcile LLM-generated patches with live developer edits, eliminating structural merge conflicts
• Implemented a distributed autonomous repair pipeline with build/test failure detection, replay-loop reasoning, and a Self-Repair Agent coordinating IDE ↔ backend execution
• Improved autonomous repair performance to 46% SWE-bench Verified and 37% Lite by enhancing agentic self-repair and failure recovery strategies

AI Researcher

• Developed a multi-agent LLM framework for large-scale citation screening using Chain-of-Thought, PICOS evaluation, and LLM-as-a-Judge, achieving 99.5% sensitivity and 87.9% specificity across 15 systematic reviews (~150K abstracts)
• Led collaboration with medical researchers; primary contributor to manuscript under review at NEJM AI (2025) demonstrating scalable agentic automation for medical text mining

Languages
Python, Kotlin, Java, TypeScript/JavaScript, C#, Swift, SQL, R

Systems & Backend
Distributed systems, async architectures, FastAPI, REST APIs, Docker, Redis, MongoDB, CI/CD, Linux, multithreading & concurrency

Agentic AI & Tooling
Multi-agent workflows, LLM tooling, SWE-bench, Self-RAG, evaluation pipelines, developer-facing AI systems

• Email: zanwen.fu@duke.edu
• LinkedIn: https://www.linkedin.com/in/zanwenfu

_{Last updated: Feb 2026}