Software Engineer — Applied AI & ML Infrastructure
I build and deploy AI applications and the software around them.
My work spans model serving, distributed training, retrieval-augmented generation, and backend systems. Previously, I worked on reinforcement-learning infrastructure as an AI Software Engineering Intern at InstaDeep and computer-vision rendering pipelines as a Software Engineering Intern at Prophesee. I studied software engineering at 42 Paris.
A deployed bilingual question-answering service for SEC 10-K and Korean DART filings.
- PostgreSQL/pgvector hybrid retrieval with source-linked answers and source-span citation checks.
- Retrieval evaluation and ablation runs, with token and cost limits for model calls.
- A tool-calling agent loop on the OpenAI API that answers only with checked citations, under iteration, token and cost budgets.
- FastAPI streaming backend and a Next.js interface.
A playable Gomoku with two AI engines, built as a two-person 42 project with Woolim Park.
- AlphaZero-style PyTorch self-play and arena evaluation, distributed with Ray and accelerated with C++ MCTS through pybind11.
- A separate C++ Minimax engine (alpha-beta, transposition table, killer moves, iterative deepening) that meets the project's 0.5 s average move-time limit, plus a PVS hard mode.
- C++ and FastAPI WebSocket servers; both engines run in Docker behind a Cloudflare Worker (AlphaZero on Google Cloud, Minimax on Oracle Cloud).
- I designed the hard-mode evaluation, which Woolim mostly implemented; the game UI is Woolim's.
- webserv — C++98 HTTP/1.1 server (42 team project): kqueue event loop, non-blocking sockets, CGI through fork/execve.
- cloud-1 — Pair project: Ansible roles that deploy a teammate's Docker Compose WordPress stack (Nginx TLS, MariaDB) to a remote Ubuntu host, started at boot by systemd.
- ft_services — Minikube lab: eight Alpine-based services, five behind MetalLB, Nginx TLS, Telegraf → InfluxDB → Grafana.
- kfs — Bootable i386 kernel with a partner: GDT/IDT and IRQs, PS/2 keyboard, a small shell, freestanding libc, ISO/QEMU build.
| Area | Technologies |
|---|---|
| Programming | Python, C++, TypeScript, SQL |
| AI & ML Systems | PyTorch, TensorFlow Serving, Ray, RAG, Retrieval Evaluation, OpenAI API |
| Backend & Data | FastAPI, gRPC, Protobuf, PostgreSQL, pgvector, Spark |
| Cloud & Infrastructure | AWS, Google Cloud (GCP), Docker, Kubernetes, Linux |
| Tooling & Delivery | Claude Code, Codex, ChatGPT, MCP, GitHub Actions |
| Web | Next.js, WebSocket |
AWS Certified Solutions Architect – Associate (Mar 2026) · AWS Certified AI Practitioner (Jan 2026)
Based in Paris, France. Open to software engineering, applied AI, ML infrastructure, and backend opportunities in France and across Europe.