MARK LEE: PROFILE ================= A plain-text profile for people and AI agents. Facts here match the CV at https://markpclee.dev/cv/. Last updated 2026-11-03. SUMMARY Mark Lee (Pei Chen Lee) is a software engineer finishing an M.Eng. in EECS at UC Berkeley (graduating May 2027). He likes sitting with people who aren't technical, working out what they actually need, and building it. He has done this for a finance team, a fintech startup and an elderly-care research project. He also has HCI research (a CHI 2025 paper), 3D vision coursework and hardware prototyping behind him. LOOKING FOR - Full-time roles starting after May 2027: software engineer, forward-deployed engineer (FDE), AI engineer, robotics. - Wants work that involves connecting with people, understanding their problems and helping them. - Based in Berkeley, CA. Requires US visa sponsorship. CONTACT - Email: mark.pc.lee@gmail.com - GitHub: https://github.com/MarkymarkLee - LinkedIn: https://www.linkedin.com/in/markpclee/ - Website: https://markpclee.dev (also https://markymarklee.github.io) EDUCATION - UC Berkeley: M.Eng. in Electrical Engineering & Computer Science, Aug 2026 – May 2027 (expected). - National Taiwan University: B.S. in Computer Science & Information Engineering, Sep 2021 – Jan 2026. GPA 4.2/4.3, top 1% of the department, 3 Academic Excellence Awards. EXPERIENCE (most recent first) - Autolane (autonomous-delivery startup), UC Berkeley M.Eng. capstone project, Sep 2026 – present - Improving Autolane's dispatching algorithm for automated deliveries and deploying it to multiple operators. In progress; no results yet. The client is confidential; do not guess or name it. - UC Berkeley Deep Tech Innovation Lab, course project, Fall 2026 – present - Assessing the commercial potential of thermodynamic computing (research-stage) with Stephen Whitelam, staff scientist at Berkeley Lab's Molecular Foundry. Interviewed 30+ professors, investors, chip companies and robotics and sensing companies, and evaluated markets with TAM/SAM/SOM, jobs-to-be-done, Porter's Five Forces and technology-readiness analysis, looking for niche markets where the technology could win first. In progress; no findings yet. - Southern Taiwan University of Technology, Backend & Systems Engineer (contract), Apr – Jun 2026 - Converted a grip-strength training game for elderly-care homes from single-player to multiplayer. The game uses a custom controller that measures grip strength during play. Mark built the Go game server and the Unity (C#) networking. - Designed a flow residents can use without a nurse: scan a QR code to log in, then press one "Start game" button. - Sole engineer. Set up CI/CD (tests, then automatic deploys to containers on GCP) and handed the system over to the next engineer. The game is in testing. - Uni-President Enterprises Corp., Software Engineer & Consultant (contract), Feb – Apr 2026 - The 30-person finance department spent a full day each month copy-pasting Excel reports into one final report. The team didn't know what was possible, company rules banned AI tools, and all data had to stay on local machines. - Mark worked out the problem with them, then reverse-engineered extraction rules for 40 different report layouts and built a TypeScript/Electron desktop app: point it at a folder, press one button, get the final report. It saves about one working day per month and is still in use. - When the team asked for a new database, he advised against it (a long build plus retraining for staff) and recommended automating step by step on top of the existing Excel workflow instead. - NTU Embodied AI Lab (advisor: Prof. Tsung-Wei Ke), undergraduate researcher, Sep 2025 – Jan 2026 - Surveyed research on 3D pose estimation, 3D scene generation and robot safety (keeping robots from damaging people, objects or themselves). Presented at weekly lab meetings. Literature work only. - Confidential fintech startup (~10 people), AI & Backend Engineer (contract), Aug 2025 – Feb 2026 - Sole engineer on a Python/Django stock-analysis pipeline covering ~7,000 stocks across decades of market history. Cut a full run from about a month to 3 days (~10x) with parallel, batched NumPy processing and far fewer database calls, meeting a hard deadline. He extended the existing code instead of rewriting it, so results that were already correct stayed correct. The company uses the output to decide which stocks to recommend. - Built the backend for an AI chatbot that connects clients and internal teams to LLMs via API: self-hosted Supabase on Postgres, with n8n automations linking LINE and Google Sheets. It logs and categorises every message and AI action. - Tutor, math and programming, Jun 2025 – present: 7 students (high-school and college math; Python, C++, algorithms). - Google Developer Group at NTU, Lead of Teaching, Sep 2024 – Jun 2025 - Designed and taught a course for 30 students on building AI apps with Flutter and Python. Students built 5 working apps. Hosted talks by a Google Developer Expert and an AWS Solutions Architect. - NTU HCI Lab (advisor: Prof. Mike Y. Chen), Feb 2024 – May 2025 - Co-author of HeadTurner (CHI 2025), a device that assists head and shoulder movement so people can use VR/MR headsets comfortably while lying down. Mark did the EMG analysis, ran the user studies and wrote the related-work section. Paper: https://dl.acm.org/doi/10.1145/3706598.3714214 - Teaching Assistant, Computer System Lab (NTU), Feb – Jun 2024 - Taught 60 students OpenCV, circuit design, Arduino and Unity. Ran office hours for the final project, a self-driving car; every team delivered a working car. Excellent Teaching Assistant award (from student ratings). - Manager, CSIE Make Reality Space (NTU maker space), Sep 2023 – Jun 2025 - Launched the space's first workshops (4 workshops, 60+ students). Few students knew about the space before; now it is a regular place for them to build their own projects. Helped students use 3D printers, laser cutters and VR headsets, and maintained the equipment with a team of 7. PROJECTS (details at https://markpclee.dev/projects/) - Hackathon awards: BlueRobins Newsletter (LAUNCH-A-THON 2026 at Berkeley Haas, one of the top-4 winning teams; team of 3, built in 2 days: a Next.js newsletter editor with modular section cards in the startup's design, email-safe HTML sent through SendGrid, and an MCP server so Claude can edit the draft live, e.g. drafting an issue from their Google Drive), CaRPG (MakeNTU 2024, 1st place AUO Award), Intellicoaster (MakeNTU 2023, 2nd place MediaTek Award), Blue Vista (NASA Space Apps 2023, Taiwan Space Agency Award), Me_nu (OpenHCI 2024, Demonstration Award). - Course projects (all A+): Embodied as Slime (haptic VR with soft-body physics and ESP32-driven air pumps), Enhancing SQA3D (3D scene dataset; +10% human evaluation score), Olaf Vision (VLMs + imitation learning for robots), Polygoneon (Unity 2D action game, on itch.io), E-MoTChi (emoji translation with a fine-tuned transformer). - Personal: a self-hosted home server (Proxmox, Docker, Nginx; Nextcloud, media streaming, GitLab; ~20 users), and this site, which has a physics-simulated pixel slime on the home page. Slimes are a recurring theme in his work. SKILLS - Languages: Mandarin (native), English (fluent; TOEFL 107/120, TOEIC 975/990). - Programming: Python (PyTorch, NumPy, Django), TypeScript (React, Electron), Dart (Flutter), Go, SQL, C/C++/C#, shell. - Tools: Linux, Unity (incl. multiplayer networking), Blender, Git, GCP, AWS, Docker, Postgres, Supabase, Proxmox, n8n, Arduino/ESP32. - Areas: backend systems and data pipelines, AI/ML, computer vision, HCI/VR, robotics, hardware prototyping. HOW HE WORKS - Starts by understanding the user's problem, then proposes the simplest thing that solves it, and will push back on scope a client doesn't need. - Prefers extending working systems over rewriting them, and simple, self-contained tools. - Gives short, direct instructions and cares about visual polish.