AI Engineer with 1+ years in Machine Learning & Robotics
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Software engineer with strong Python fundamentals and hands-on experience building data processing pipelines, distributed back-end services, and containerized infrastructure for production environments. Built Kafka-based event-driven microservices with SQL-backed persistence and Maven-tested reliability, alongside LLM-driven automation pipelines and Docker/Kubernetes-deployed systems across research and industry settings. Detail-oriented and collaborative, with a track record of writing clean, well-tested code and supporting data-intensive workflows alongside researchers and engineers.
University of California, Berkeley (UCB)
Master of Engineering · Electrical Engineering & Computer Science
August 1, 2025 – June 30, 2026
National Taiwan University (NTU)
Bachelor of Science · Electrical Engineering
August 1, 2020 – June 30, 2025
TrendForce Corporation
Industry/Lab Researcher
October 1, 2024 – June 1, 2025
Taiwan
Pegatron Corporation
Intern Software Engineer
July 1, 2023 – August 1, 2023
Taiwan
IADIY Photonics
Intern Software Engineer Webpage Developer
July 1, 2020 – August 1, 2021
Taiwan
JPMorgan Chase Forage Software Engineering Simulation
April 1, 2026 – April 1, 2026
Built a Spring Boot microservice to consume and deserialize high-volume Kafka transaction messages with configurable topics. Implemented validation and persistence with Spring Data JPA and H2, modeling entities and balance updates across User records. Connected the service to an external REST Incentive API using RestTemplate and exposed a balance-query endpoint via a Spring controller, returning JSON responses while maintaining clean architectural boundaries.
View ProjectRisk-Triggered Latent-Space Corrections for Near-Failure Recovery
February 1, 2026 – May 1, 2026
Designed a decoupled correction architecture for Soft Actor-Critic (SAC): a lightweight MLP applies additive latent corrections Az to the actor's hidden state, gated by a geometric near-failure trigger to only 2-12% of steps at convergence. Applied stop-gradient on the actor encoder to prevent co-adaptation and preserve independent base actor capability. Evaluated across 4 Meta-World manipulation tasks; improved mean final success rate by +57pp on push-v3 (32.5%→89.5%) and +27.5pp on pick-place-v3 (66.0%→93.5%), achieving full seed reliability (4/4) versus 0/4 under concurrent training.
View ProjectLangGraph ML Research Buddy
January 1, 2026 – June 1, 2026
Built a stateful six-agent LangGraph pipeline (planner, log analyzer, claim checker, reward-hacking auditor, critic, writer) with conditional edges, bounded retry/reflection loops, and a human-in-the-loop checkpoint for ML experiment review. Backed by SqliteSaver for durable execution, Pydantic structured outputs per node, a custom BaseCallbackHandler tracing LLM calls to per-run artifacts, and a FastAPI/SSE streaming interface; provider-agnostic across Anthropic, OpenAI, and Ollama.
View ProjectRAG-Forge: Modular RAG Pipeline Framework
January 1, 2026 – June 1, 2026
Designed and implemented five end-to-end RAG pipelines from naive dense retrieval to graph-based, agentic, and LangChain LCEL architectures, with a provider-agnostic interface supporting Anthropic API, OpenAI API, and Ollama. Implemented hybrid retrieval (BM25+dense search, RRF fusion, cross-encoder reranking, HyDE).
View ProjectAI-driven Personalized Assisted Device
August 1, 2025 – May 1, 2026
Built a 3-phase LLM-guided optimization pipeline over a 15-dimensional design space, converging to a personalized assistive device design in as few as 2 iterations and achieving 83% overall convergence rate across 12 runs and 4 participants. Designed task interfaces across 8 simulated grasping environments, enabling evaluation of grip force and tasks success rates. Optimized designs reduced required grip force to a mean of 2.5 Nm, with 7 of 8 tasks achievable at or below 1.7 Nm.
View ProjectIMU-Controlled Quadruped Robot
August 1, 2025 – December 1, 2025
Implemented a 4-stage locomotion pipeline (sinusoidal trajectory, planar IK, gait coordination, motor dispatch) supporting 5 gaits including trot, walk, skid-steer turning, and jump, with stride up to 50 mm and forward velocity of 17 cm/s (>1 body length/s). Debugged Raspberry Pi 4 to Dynamixel RX-24F UART communication, synchronizing all 8 joints at 100 Hz with 0.29° per-tick resolution over a 1 Mbps daisy-chained bus. Built an FSM command dispatcher routing IMU/EMG outputs to joint trajectories; evaluated SVM and DNN classifiers on 8-channel SEMG at 250 Hz, reaching up to 40% gesture accuracy with IMU-threshold control deployed for the final demo.
Vision Fairness Analysis - CNN vs Vision Transformer
August 1, 2025 – December 1, 2025
Designed a fairness study comparing U-Net encoders (ResNet50 vs. FastViT, ImageNet-1k pretrained) on 4× super-resolution and racial classification across FairFace (108k images) and RFW datasets, simulating 1:20 minority-to-majority imbalance per race. Fine-tuned via staged layer unfreezing using perceptual and cross-entropy losses; ResNet50 achieved +5.79 dB higher PSNR than FastViT, with both models showing ≤0.05 dB per-race variance, confirming super-resolution is race-agnostic.
View ProjectSmart Parking
August 1, 2023 – January 1, 2023
Collaborated with a teammate to develop a smart driving system, featuring real-time notifications and auto-parking. Integrated PiRacer Pro and STM32 board to boost system capabilities with analytics and imitation learning.
View ProjectDynamic-SUPERB Phase-2: An Open Benchmark for Evaluation of Spoken LMs with 180+ Tasks
ICLR
January 1, 2025 – Present
LLM Discussion: Enhancing the Creativity of LLMs via Discussion Framework and Role-Play
COLM
January 1, 2024 – Present
Reviewer
2024 IEEE Spoken Language Technology Workshop (SLT)
January 1, 2024 – Present
Cultural Fit Analysis
The candidate's diverse project portfolio, spanning academic research, personal initiatives, and a professional simulation, indicates a broad interest in various AI/ML applications and a strong drive for continuous learning. The academic background from top-tier universities (UC Berkeley, National Taiwan University) and involvement in research labs suggest a collaborative and intellectually curious individual. The experience as a reviewer for a technical workshop further highlights engagement with the broader scientific community. The blend of theoretical knowledge and practical application in projects like 'Risk-Triggered Latent-Space Corrections' and 'AI-driven Personalized Assisted Device' shows a well-rounded profile suitable for a dynamic R&D environment.
Soft Skills & Operational Fit
The candidate demonstrates strong problem-solving skills through complex project implementations and debugging efforts (e.g., Raspberry Pi communication). Collaboration is evident in team projects and research roles. The detailed project descriptions suggest a methodical approach to design, evaluation, and optimization. The candidate's involvement in academic research and publications indicates a proactive and inquisitive mindset, which aligns well with an AI Engineer role requiring continuous learning and innovation.