Zhoujing CHEN

Gofo France, Paris

• Built a Python + CPLEX/DOcplex postal-code districting optimizer using MIP, adjacency graphs, and flow/boundary-penalty formulations to create contiguous, compact, workload-balanced DSP territories under excluded, bound, and locked-code constraints.
• Developed a Python + OSRM DSP park-and-walk route planner with service-stop clustering, parking-candidate snapping, and TSP solving (heuristic fallback + exact CPLEX for smaller instances) to improve dense urban last-mile routing.
• Built a Python + OSRM + CPLEX mixed-fleet linehaul/distribution CVRP with pallet capacities, trip-cost pricing, route-duration limits, and secondary-depot scenario evaluation to compare direct-ship and hub-assisted network designs.

First Moment, Shanghai

• Built a minute-level BTCUSDT perpetuals RL trading framework with separate train/validation and a strict out-of-sample holdout; evaluated under next-bar VWAP execution with realistic transaction costs.
• Designed a continuous-action policy with standardized time series features; mapped target position to fixed notional partial adjustments to cap turnover and suppress noise trades.
• Proposed a reduced-clip schedule for PPO with momentum-based pretraining; implemented baselines (DQN, DDPG, Bollinger, momentum) and observed superior risk-adjusted performance on the holdout.

Group R3 – CentraleSupélec, Gif-sur-Yvette

• Built a reproducible CPLEX/OPL environment and modeling guide; implemented a baseline intermodal MILP and a synchromodal extension where flexible services activate only after fixed capacity is saturated, to quantify resilience benefits.
• Extended the models to a time-expanded network with full cost accounting (timetables, handling/warehousing, carbon tax) and disruption-adjusted truck travel times, enabling rapid replanning under sudden outages.
• Prepared a scenario suite (baseline demand, capacity reductions, link outages—single-node to regional cascade) and compared integer-flow vs. relaxed formulations with abandonment penalties to assess trade-offs in solution quality and runtime.

Merit Interactive Co., Ltd., Hangzhou

• Developed and maintained large-scale backend and data processing services using Java, Spring Boot, Spark, Amazon S3, and ClickHouse on Kubernetes.
• Participated in the full SDLC from requirements analysis to deployment, delivering multiple feature iterations on schedule.
• Built and fixed RESTful APIs with clear controller and service layers; produced internal API documentation.

LOGEFI SERVICES, Paris

• Automated end-to-end French VAT (TVA) workflow on impots.gouv.fr (login, declaration/payment, receipts/status), serving 20k+ client entities.
• Implemented with Selenium, openpyxl, PyMuPDF, pandas, and Tkinter. Packaged PyInstaller apps for non-dev use.
• Ongoing maintenance (Sep 2023 – Present): multiprocessing crawler and CNN-based automated CAPTCHA recognition.