Connecting AI/ML, Woks, Wrenches & Software Systems
I'm Noah Han. I build NLP/LLM systems and recommendation engines in the Bay Area, but my engineering mindset was forged in high-heat restaurant kitchens, refined under intense academic environments, and tested on real-world plumbing and automotive systems.
Culinary Mastery & Business Operations
Before entering college, my life was anchored in our family's busy restaurant. Operating within a family business means taking on every role imaginable. As a certified chef, I mastered high-volume Chinese culinary techniques, outperforming tier-2 professional chefs under heavy rush hours.
Beyond cooking, I worked closely with management to optimize operations, streamline supply costs, and restructure kitchen workflows. Our primary target was clear: drive profitability and operational efficiency. The ability to analyze bottlenecks in a high-stress kitchen and translate those insights into business returns laid the groundwork for my career in software, where optimizing performance directly impacts operational cost.
Systematic Learning & High-Stakes Exams
Transitioning into higher education culminated in preparing for the highly competitive National Postgraduate Entrance Examination (NSEEE) in China. This massive, unified nationwide exam sees millions of candidates compete annually, with admission determined by individual universities and programs. Targeting Peking University's School of Software and Microelectronics, I competed against thousands of candidates for approximately 600 admission slots. My structured preparation systems yielded excellent results: I ranked in the top 10 among the 600+ admitted students to the School of Software and Microelectronics at PKU.
To achieve this, I formulated and refined personalized study frameworks based on the Feynman Technique and active recall. This method of breaking down massive, complex, and ambiguous subjects into modular, highly digestible logical structures is the exact same strategy I use today when diving into unfamiliar software repositories or conducting technical root-cause analyses.
NLP, LLMs & Recommendation Engines
My core technical focus lies at the intersection of natural language processing, large language models, and recommendation systems. I design and build intelligent systems that understand, generate, and personalize content at scale.
I have hands-on experience building RAG (Retrieval-Augmented Generation) pipelines that combine vector databases with LLMs for accurate, context-aware knowledge retrieval. On the recommendation side, I've worked on collaborative filtering, content-based models, and hybrid architectures that deliver real-time personalization. My toolkit includes Python, PyTorch, vector databases, and modern ML deployment infrastructure.
What excites me most is bridging the gap between cutting-edge ML research and production-grade engineering — turning models into reliable, observable, and scalable services that solve real user problems.
Utility Retrofits & Wrench Turning
My passion for engineering spills directly into physical structures. I believe if you own a system, you should understand how to disassemble, repair, and optimize it.
When my home's traditional tank water heater needed replacement, I bypassed commercial contractors and executed a complete conversion to a high-efficiency gas tankless water heater myself—managing gas pressure calibrations, water flow piping, and venting regulations. Alongside this, I retrofitted my home with custom-wired smart switches, lighting control grids, and secure smart locks. On the road, I apply this same hardware curiosity to vehicle electrical systems, custom dashboard cameras, and 12V wiring diagnostics.
Philosophies
My Four Pillars of Composure
1. Mise en Place
The culinary discipline of organizing ingredients and tools before starting. In software engineering, this is system architecture planning—ensuring environment settings, data pipelines, and interfaces are defined before writing the first line of code.
2. Data-First Thinking
In ML engineering, the model is only as good as the data. I apply rigorous data exploration, feature engineering, and evaluation frameworks before jumping to architecture decisions — letting evidence guide the solution rather than intuition.
3. Martial Focus
Years of martial arts conditioning built strong focus. During intensive training, I developed intermediate mastery of traditional skills including Vajra Iron Palm (Da Li Jin Gang Zhang for brick and walnut breaking), Cotton Belly (Mian Hua Du for core abdominal control), and sound-localization reflexes. Though long-term absence from active conditioning has reduced these physical techniques, the elevated hand bone density and, more importantly, the psychological composure under extreme pressure remain intact.
4. Physical Feedback
Physical engineering has immediate, non-negotiable feedback loops—a gas leak or a plumbing drip is instantly visible. I bring this standard of strict quality assurance and rigorous edge-case testing back to my software and ML systems.