resume(1) — curriculum vitae

Recent Software Engineering graduate from UC Irvine with solid skills in Python, Java, and SQL. Strong understanding of data structures, algorithms, and design principles. Experienced in software development, testing, and project management through academic projects. Eager to contribute in cross-functional teams and build scalable, clean, and efficient software solutions.

• University of California, Irvine

  [2021-09 → 2025-03]

  B.S. in Software Engineering

  • · GPA: 3.459
  • · [coursework] Design and Analysis of Algorithms, Internet Applications Engineering, Data Structure Implementation and Analysis, Software Design: Structure and Implementation, Software Design: Applications, Human-Computer Interaction, Principles of Operating Systems, Information Retrieval, Project in Databases and Web Applications, Software Testing, Analysis, and Quality Assurance

1. §03.01 ./ascension-ai/

   [2025-11 → present]

   • · Built end-to-end RL pipeline with a 717-d structured observation encoder and 134-action masked policy for a real-time game environment.
   • · Implemented PPO from scratch with clipped objective, GAE advantages, target-KL early stopping, and a BC anchor loss to prevent catastrophic forgetting during fine-tuning.
   • · Made deck-building a learned skill: added a per-card deck observation vector and moved card removal/upgrade from a heuristic into the RL policy with a potential-based deck-quality reward, migrating the trained model 585→717 via behavior-preserving warm transfer + a BC anchor.
   • · Designed parallel rollout architecture: 4–8 concurrent worker processes feeding a central offline trainer via checkpoint-tagged .npz files, with stale-rollout rejection.
   • · Deployed the stack headless on a GPU-less GCP spot VM via a one-shot installer (8 instances under per-worker Xvfb + software OpenGL), debugging display contention, LWJGL native-extraction SIGSEGV races, Java 8 mod loading, a 10 s mod-handshake timeout, and a silent JVM heap OOM.
   • · Built self-healing, session-independent training ops: a per-worker watchdog + a VM cron that continuously auto-resumes training (with an auditable heartbeat) + a Cloud Scheduler job that restarts the spot VM after preemption — recovering in ~15–25 min with no human.

2. §03.02 ./bitlink/

   [2024-03 → 2024-06]

   • · Shipped a full-stack Angular + Node.js social platform with real-time messaging and social interactions.
   • · Implemented secure auth with Passport.js and BCrypt; designed a responsive UI in Angular + TypeScript.

3. §03.03 ./capstone-archive/

   [2024-01 → 2024-06]

   • · Built a React + Django web app for hosting student capstone projects.
   • · Integrated Firebase Authentication and a MySQL8 database; ran the team on Agile workflows via Jira and Discord.

4. §03.04 ./search-engine/

   [2023-02 → 2023-03]

   • · Engineered a Python search engine indexing 56,000+ web pages.
   • · Reduced average query response time by 35% via custom tokenization and algorithm optimization.

[languages]

java, python, c, c++, javascript, typescript

[web]

html, css, angular, react, django

[data]

sql, mongodb, firebase

[ml/ai]

reinforcement learning, ppo, behavior cloning, gae, actor-critic, action masking, reward shaping, pytorch, gymnasium

[engineering]

distributed systems, ipc, parallel workers, atomic checkpointing, crash recovery, software testing, algorithms, design patterns, api integrations

[methods]

agile/scrum, devops practices, project management