Our Work
A sampling of our projects across R&D and engineering.
Agentic Document Processing
AI Startup
- Implemented memory-empowered AI agent with significantly higher accuracy vs. context-window-only baseline
- Built line-by-line file editing agent, dramatically reducing token costs
- Constructed high-speed PDF-to-HTML pipeline via intelligent parallel processing
AI-Powered Chip Design
Chip Startup
- Advised on the design and implementation of a microfluidic cooling system for high-performance chips
- Prepared a report on the semiconductor supply chain and unit economics
- Built an AI-powered chip design and verification tool to accelerate the development cycle
Semantic Search Engine
Investment Firm
- Built vector database over internal datasets for fast retrieval by investment thesis
- Designed intuitive search interface for non-technical analysts
- Migrated legacy data into modern infrastructure
AI Model Training Platform
AI Startup
- Built Kubernetes-orchestrated fine-tuning platform with on-demand AWS GPU provisioning
- Dockerized training jobs with auto-scaling infrastructure for cost-efficient compute
- Added live inference engine for real-time model evaluation during development
AI Hardware Due Diligence
Venture Capital Firm
- Advised on the infrastructure development for AI workloads, including GPU provisioning and high-performance computing
- Advised on the development cycle of ASICs and specialized AI accelerators such as photonic processors
- Prepared a report on the specifications and performance of AI accelerators of a startup for due diligence
Deep-Tech Due Diligence
Venture Capital Firm
- Advised on the computer system and networking infrastructure for AI workloads
- Advised on the technical viability of a photonic chip-based oscillator for AI processors
- Prepared a report on the specifications and performance of the photonic oscillator for due diligence
AI Model Workflows
Autonomous Vehicle Company
- Shipped safety-critical behavior model improvements across features, architecture, data pipelines, and eval loops
- Hardened release workflows with higher label/data quality, regression standards, and automated pre-merge checks
- Scaled ML evaluation via standardized test suites, metric gates, and labeling feedback—cut iteration and release analysis time >5×
AI Developer Enablement
AI Startup
- Delivered a 6-week full-stack AI builder curriculum for 125+ developers, from fundamentals to deployment
- Built project-based build sprints and live debugging drills, helping teams ship prototypes >2× faster
- Scaled a tutorial and debugging content engine, driving 2,000+ YouTube views and 200+ X followers