Hi, I'm Richard Shan
I'm a
I build intelligent systems: generative AI, RAG pipelines, and embedded platforms. Passionate about LLM evaluation, retrieval architecture, and machine learning that matters.
About Me
I'm a systems-focused engineer working across generative AI, machine learning evaluation, and embedded computing. My work focuses on generative AI assurance and evaluation and building real-world applications.
With a strong computer science foundation and an eye for systems design, I enjoy building products with tangible impacts. From retrieval pipelines to embedded AI for accessibility, I bring a builder's mindset to complex technical challenges to create intelligent systems that matter.
When I'm not coding, you'll find me translating Latin and Greek poetry, working in venture capital, and playing ultimate frisbee. I'm passionate about accessibility, education, and the intersection of technology and social impact, and I'm always looking for ways to apply technical skills beyond the screen.
Recently, I led the development of DynaRAG, a dynamic framework for retrieval-augmented generation that optimizes chunking strategies for context-aware LLM performance; and founded Brailliant, a modular refreshable Braille translator and display powered by solenoids and embedded intelligence. These projects reflect my focus on building novel end-to-end systems that combine AI research with practical design to enhance human-AI interaction.
Work Experience
Translating research into production systems with end-to-end ownership across hardware, software, and evaluation.
Founder & CEO
Brailliant
Building affordable accessibility hardware.
- Developed modular refreshable Braille translator and display powered by solenoids and embedded intelligence
- Raised $10,000+ in pre-seed funding from accessibility-focused investors
- Established relationships with key stakeholders in the visually impaired community
Generative AI Scientist
The MITRE Corporation
Researching AI assurance for national-security applications.
- Developed mechanistic interpretability frameworks and 6 quantitative metrics for analyzing LLM internals
- Discovered novel feature-controlled LLM reasoning modalities
- Proved LLM features tend to fracture into domain experts at high task complexities
- Presented findings to Congressional and DoD staff
AI Alignment Research Intern
Stanford University School of Medicine
Evaluating AI safety in healthcare applications.
- Designed and implemented a retrieval-augmented generation pipeline for clinical discharge anomaly detection
- Conducted red-teaming and adversarial testing on LLMs in clinical contexts
- Won 2nd Place in Stanford-internal hackathon for modelling chest x-ray pneumonia detection
Explore More
Dive deeper into detailed write-ups covering research insights and end-to-end builds.
Systems, Evaluation, Assurance
Learn about my published research on retrieval architectures, generative AI assurance, computational linguistics, and more.
View research portfolio↗Builds, Deployments, Experiments
Come along for my journey through interactive hardware, embedded systems, and edge AI.
Explore project library↗Skills & Technologies
GenAI Evaluation
Data Analysis
Retrieval-Augmented Generation
Vector Stores
Agentic AI
Embedded Systems Engineering