Neil Lu

building ai agents in/for hardware labs context management, compression design, claude skills, on tauri

Prototyped multi-agent compliance automation, with supervisor-delegation orchestration, knowledge-base management, MCP tool use, and human-interveneable safety mechanisms.

With B2B users already onboarded and NT$1.2 million secured, we bridge employers and early-career job seekers, funded by Taiwan's Ministry of Education and Hotai Motor. • Architected full-stack platform (Next.js, Firebase, and Go), implementing security rules, CI/CD, RWD/PWA, and SEO optimizations. • Developed fuzzy search to search in 550k+ companies in < 50ms, and chat forum with realtime WebSocket streaming & Cloud Functions. • Built data schema, caching, & permission frameworks for multi-provider OAuth, post feeds, notification, admin modules, etc. • Co-led 7 (2 engineers); drove growth via 1,000+ cold emails & social (200k+ views → 500+ signups, 100+ user posts in 3 months). • Secured ~US$40k+ pre-seed funding, won startup competitions, and earned press coverage in multiple outlets after my joining.

In charge of prototyping a new test framework and implementing LLMs into workflows. • Integrated LLMs with auto context selection into the testing framework, enabling auto-script generation and updates. • Prototyped Playwright-based testing framework to replace original solution, was adopted for >80% of new test cases within 3 months. • Coordinated across QA/SDET to create new workflows, held workshops for the new framwork, and iterated features based on feedback. • Built and maintained features for existing e2e testing system (Python, Selenium, shell scripts, dockerfiles) with containerized deployment, ensuring compatibility on GitHub Actions, AWS Spot Instances, and local dev environments.

Spent 4 months learning how to defend a democracy. We were trained on teamwork, swimming, weaponry, emergency healthcare, urban warfare, and how to support professional soldiers.

See https://neillu.com/articles/itri In charge of a new project about brainwave (EEG) signal acquisition and processing. • Built Python framework to stream brain signals from different hardware; developed portal with HTML & JS for realtime visualization. • Optimized Python framework with parallel processing/threading to perform realtime filtering, separation, & analysis of signals. • Scripted Arduino systems to stream analog sensor data into chip and then server; analyzed & modified circuits for brain measurement. • Produced patent summaries & market reports for commercially available brain-computer interface devices and mixed reality systems.

Contracted to build this finance discussion forum for one of Taiwan's most active financial communities. • Developed a finance forum in Next JS, with PHP & MySQL as RESTful API backend, handled 6k+ signups in 2hr without crashes. • Used Cloudflare’s security platform; used Vonage, TwSMS, and LINE API to develop a multinational 2-step verification system.

Founded this platform where students can search 20 educational platforms using a single UI. 1 million visitors reached; thousands signed up; got featured on mainstream media. • Led a nonprofit team of 10 to develop, fund, and market the educational resource platform, which has reached 1.3 million user visits. • Configured Apache on Linux VM, built CI pipeline on GitHub Actions for deployment, with separation of dev and production modes. • Developed the platform with Django, MySQL, & Next JS, improving load time and SEO; multiple pages rank 1st on Google Search. • Implemented a content fuzzy-search algorithm with BERT-based models; built AI tutor with GPT-4 & retrieval-augmented generation. • Built a Chrome extension for crowdsourcing and a management system for content moderation, received 1,300+ resources from users.

Built intelligent systems with natural language understanding and common knowledge. • Developed a prototype information extraction system to extract biochemical interactions from research papers with spaCy (for NLP). • Parsed 100% of the Chinese Wikipedia for text mining to construct a common knowledge graph with Python and SQL. • Designed an error-tolerant word-rec algorithm for Chinese pharma terms, with 97%+ correctness & 300% avg. runtime improvement. • Assessed the use of pretrained word embeddings in meaning-disambiguation and visualized results with Python, HTML, CSS.

I cofounded this volunteer organization that has enabled 3,000+ hours of tutoring, serving underserved students. I also co-developed an online learning management system (LMS) and a volunteer-matching system to facilitate our work.

Gap year in 24/25 school year: Internship, and military in Taiwan's Marines

2nd Major in addition to ECE