Production experience in reinforcement learning (AlphaZero, DQN, SEED RL), ML infrastructure (gRPC, TensorFlow Serving, Ray), and cloud deployment (GCP, Docker, Kubernetes).

Professional Experience

AI Software Engineering Intern, InstaDeep

Paris, France

Mar. 24 - Sep. 24

  • Researched and implemented R-Tree spatial indexing for 3D bin-packing within a reinforcement learning training pipeline, benchmarking against brute-force baselines.
  • Integrated a SEED RL inference server using gRPC and TensorFlow Serving in C++, enabling distributed inference for 3D bin-packing agents.

Software Engineering Intern, Prophesee

Paris, France

Sep. 22 - Mar. 23

  • Researched and implemented performance enhancements and new features for ray/path tracing, improving overall rendering performance by 25%.
  • Ported the rendering pipeline across Blender versions, integrating new ray/path tracing algorithms and configuring CUDA infrastructure for the computer vision pipeline.

Data Analyst, TPMN

Seoul, South Korea

Feb. 19 - Jan. 20

  • Analyzed and optimized automated transaction processing of 5TB+ daily data, reducing processing time by 50%.
  • Designed and built an hourly automated analytics data pipeline serving the company-wide analytics environment.

Projects

Gomoku [Live]

Oct. 24 - Feb. 26

  • Trained an AlphaZero Gomoku agent using PyTorch with MCTS self-play, scaling search across four parallelism backends from sequential to Ray-distributed for near-production training.
  • Built a C++ Minimax engine with alpha-beta pruning, principal variation search, iterative deepening, transposition tables, and killer move heuristics for real-time move computation.
  • Deployed to production on Google Cloud with Docker and Cloudflare Workers routing. Built a Nuxt 3 frontend with real-time WebSocket gameplay and move evaluation.

Learn2Slither

Jan. 26

  • Built a DQN agent for Snake using PyTorch with a 24-dimensional relative state representation encoding wall, food, and body distances in four directional axes from the snake's heading.
  • Designed multi-signal reward shaping with proximity incentives, BFS-based trap detection, and starvation penalties. Fully YAML-configurable with checkpoint persistence for incremental training.

multilayer-perceptron

Jan. 24

  • Implemented a multilayer perceptron from scratch using NumPy with manual backpropagation, four gradient descent optimizers (SGD, Adam, Adagrad, RMSProp), and numerically stable softmax with cross-entropy loss.
  • Applied He initialization, mini-batch training, and early stopping for breast cancer classification on the WDBC dataset with model checkpoint persistence.

Education

42

Paris, France

Mar. 22 - Jan. 26

  • MSc (RNCP Level 7)
    • Database Architecture and Data, Dec. 24
    • Network Information Systems Architecture, Mar. 25
  • Ranked top 0.1% worldwide (50th among 40,000+ students) and top 30 in Paris (among 6,400).

42 Seoul

Seoul, South Korea

Feb. 20 - Mar. 22

  • Algorithms, systems programming (C), network architecture, and Unix administration.

The University of Oklahoma

Norman, OK, United States

Aug. 09 - Dec. 17

  • Economics