Experience

I work across machine learning systems and compilers — inference runtimes, neural architecture search, and the infrastructure that makes both go fast. Below is what I've been up to. For degree and coursework, see Education.

Industry

• Machine Learning Runtime Engineer Intern · Cerebras Systems · May 2026 – Aug. 2026
  • Co-developed a waferless simulator, enabling off-wafer inference debugging and performance benchmarking.
  • Refactored the inference metadata transmission pipeline, merging Token Metadata onto the Simulation Metadata Tensor and reducing simulation latency by 10 ms.
  • Engineered architectural support for Cerebras's second-generation inference on the Waferless Inference Simulator.
  • Added support for testing and modelling the impact of different context lengths, annotating performance into RTIR.
  • Annotated RTIR performance on scheduler cycles for state-of-the-art models on both the Wafer Engine and the simulator.
• Student Researcher · Huawei · Aug. 2025 – Dec. 2025
  • Reduced CPU core utilization and memory consumption by 10+% over static allocation through dynamic resource orchestration with runtime load-aware scheduling.
  • Increased computational efficiency of large-scale data processing workflows via UDF-level compiler optimizations with automatic vectorization and parallelization.
  • Improved stability of concurrent execution workflows under fluctuating workloads with real-time monitoring and scaling strategies, reducing execution latency.

Research

• Undergraduate Research Assistant · The Matter Lab · Apr. 2026 – present
  • Accelerating std::autodiff and adding GPU support to a deep learning framework in Rust with std::offload, supervised by Alán Aspuru-Guzik and Varinia Bernales.
• Machine Learning Engineer · Acceleration Consortium · Oct. 2025 – present
  • Researching bottom-up Neural Architecture Search (NAS) algorithms that synthesize minimum viable neural networks under tight memory budgets.
  • Developed reinforcement-learning-guided evolutionary NAS algorithms.
  • Implemented a scalable training-free evaluation pipeline using aggregated zero-cost proxies to estimate model performance, cutting evaluation time by over 90%.
• Undergraduate Research Assistant · IDs Lab, Sichuan University · Jun. 2025 – Apr. 2026
  • Researched agentic memory system architectures that enforce architectural isolation between trusted system instructions and untrusted external retrievals.
  • Developed latent-space filtering mechanisms in the vector retrieval pipeline to identify and neutralize adversarial embeddings before context integration.
  • Implemented in-context partitioning protocols during the memory write phase to prevent data poisoning and maintain long-term memory integrity.

Competitive Programming

• Co-President and team coach of the University of Toronto ICPC Club.
• Placed 12th of 52 teams at the 2026 ICPC North America Championship — awarded high honors and qualified for the 2026 ICPC World Finals.

Technical Skills

• Languages: Rust, Python, C/C++, Java, C#, SQL (Postgres), HTML/CSS, JavaScript, MATLAB
• Technologies: Linux, MongoDB, AWS, Git, CI/CD, Docker, TensorFlow, PyTorch, LLVM, MLIR, Triton, CUDA
• Frameworks: React.js, Node.js, Flask, Django, FastAPI, .NET, LangChain, LangGraph, vLLM