AI/Game Development Engineer
UC Berkeley Operations & Behavioral Analytics Lab
Sep 2025 - Present
- Modeled sequential decision-making under uncertainty in delayed-reward environments using Deep Q-Networks (DQN); improved policy learning via reward shaping, experience replay, target networks, and hyperparameter sweeps.
- Built an experiment-grade simulation platform in SvelteKit with Firebase (auth, realtime streaming) to collect event-level trajectories and state/action/reward logs for policy evaluation and behavioral analysis.
- Quantified behavioral effects in human–agent interaction data using controlled comparisons and uncertainty estimates (confidence intervals / bootstrap) over key outcome metrics; presented with Prof. Park Sinchaisri at Stanford and CMU.
Python
PyTorch
JavaScript
TypeScript
FirebaseData Engineering Intern
Shenwan Hongyuan
Jan 2025 - Feb 2025
- Owned market-data QA + reconciliation for trading pipelines; implemented schema/null/outlier and symbol/timestamp alignment checks with researchers; optimized Pandas/NumPy/Spark transforms for 20% faster runtime.
- Automated and hardened ETL for time-series market feeds (Python/SQL/Airflow): idempotent jobs, backfill-safe runs, and dependency-managed DAGs; improved end-to-end ingestion/processing throughput by 10%.
- Tuned PostgreSQL performance for production queries (partition-aware access patterns, query-plan optimization), cutting query latency by 30%; added Redis caching for hot datasets and containerized services with Docker for deployment.
Python
SQL
Pandas
NumPyPostgreSQL
DockerResearch Intern
MIT Pentelute Lab
Feb 2023 - Apr 2024
- Developed a stochastic generative modeling + optimization stack for peptide candidates using diffusion models (sampling) and deep reinforcement learning (policy/value optimization) in PyTorch with vectorized preprocessing in NumPy/Pandas.
- Designed the RL reward as a multi-objective utility: maximize predicted affinity and developability scores while penalizing constraint violations (aggregation/toxicity heuristics), turning peptide design into a constrained optimization loop.
- Improved reproducibility by tracking configs, data lineage, training (loss/grad norms), RL dynamics (reward curves/Q estimates), and generation KPIs (top-k scores, diversity/uniqueness, sequence-level stats) across sensitivity analyses.
- Published a research paper with Dr. Vladimir Akhmetov on AI-assisted Alzheimer's drug design (DOI: 10.36838/v7i4.5).
PythonPyTorchNumPyPandasAcademic Researcher
Oxford University
Dec 2023 - Jan 2024
- Ranked 1/45 in a research team modeling biochemical systems through stochastic processes and algorithmic complexity.
- Implemented Monte Carlo simulation and time-series / nonlinear dynamical systems models in Python, C++, and MATLAB, including parameter sweeps and stability/sensitivity analysis of molecular interaction dynamics.
- Built high-dimensional statistical learning pipelines in NumPy/Pandas/scikit-learn, applying PCA (numerical linear algebra), k-means (unsupervised clustering), and Random Forest for candidate scoring and uncertainty/risk-style assessment.
Python
C++NumPyPandas
scikit-learn2024
New Zealand Chemistry Olympiad – 2× Gold Medal, 1× Silver Medal; National Exam Full Score Achiever2024
International Chemistry Olympiad Qualifier; New Zealand Chemistry Olympiad National Team Member2024
New Zealand Mathematical Olympiad Camp Member; New Zealand Senior Maths Competition 1st Place2024
British Biology Olympiad Gold Medal; British Physics Olympiad Gold Medal; UK Chemistry Olympiad Gold Medal2024
New Zealand Young Physicists' Tournament 2nd Place2023
New Zealand Young Physicists' Tournament 3rd Place2023
New Zealand Physics and Maths Competition 7th Place; International Young Physicists' Tournament Qualifier2021
Australian Mathematical Competition – 1× Medal (1/5000) & Top in Auckland City, 3× High DistinctionContact Form
Please contact me directly at nchen06(at)berkeley.edu or drop your info here.