Passionate technology enthusiast and highly adaptable software engineer with over 20 years of experience spanning full-stack development, distributed systems, and cutting-edge AI infrastructure. Driven by a deep curiosity for modern software and hardware, I excel at identifying optimal implementation paths for complex ideas and delivering robust, scalable solutions. Rather than being defined by specific programming languages, I leverage a strong engineering intuition to navigate rapid technological shifts. Whether designing multi-agent workflows, authoring core database extensions, or exploring 3D printing and hardware integration, I thrive on solving difficult engineering challenges and continuously pushing the boundaries of automation and product development.
Intelligent InternetSenior Machine Learning Engineer
Oct 2024 – Present
Reference: Manager, Ju Huo, Canada
Spearheaded the design and development of multiple AI Agent harness systems and RAG framework. Managed end-to-end ML lifecycles, taking full accountability from curating and cleaning large-scale datasets to model training and final product deployment.
Authored and currently maintain Evoke, the world’s fastest PostgreSQL BM25 extension, and trained a bespoke SAE encoder model on top of it to enable advanced Sparse Semantic Retrieval (SSR).
Based on this continuously converging lexical and semantic unified index, I built an AI agent called Commons on top of it. Commons indexes the latest papers from arXiv and PubMed in real time. Users can query the agent for the latest papers at any time, discuss a specific paper in depth, and ask what research ideas may be relevant to their own work and what insights they might gain. So far, Commons has indexed 3M aeXiv papers and 8M PubMed papers, and has achieved millisecond-level hybrid search.
Dream LensAI Developer
Oct 2023 – Oct 2024
Engineered distributed inference pipelines for image and video generation. Leveraged PyTorch and TensorFlow to optimize, and deploy generative AI, facial recognition, feature extraction, and human posture estimation models. Continuously evaluated and integrated modern technology stacks to accelerate product delivery and meet stringent project requirements.
EscapeSide Project
Oct 2026 – Present
A physics-based 3D driving game developed entirely using an automated AI agent harness. Collaborated with specialized coding agents to transition from concept to a high-quality 3D prototype, tackling complex engineering challenges like 3D visual convergence within an agentic environment across diverse terrains, weather conditions, and vehicle physics.
TensorTrace.netSide Project
Oct 2026 – Present
An autonomous, AI-curated community platform that continuously iterates a database of AI models, news, papers, and inference providers. Utilizes vector databases and SSR matching to process massive data streams entirely through multi-agent collaboration, designing a new paradigm for automated website generation and information integration. This project also uses the previously mentioned Evoke engine, supported by a unified index of lexical and semantic information.