2024–2025
Enterprise GenAI Orchestration
Scaled workflow automation and anomaly detection across industries. Agent layers, tool use, and human-in-the-loop. High conversion on enterprise pilots.
AI-Native Builder
David (Runyu) He
CMU MISM · UCLA Psychology & Computing. Building agentic systems, agent memory, harness engineering, and LLM evaluation frameworks. From orchestration layers to vector retrieval—shipping AI that scales.
David's Claw
A conversational concierge—ask about schedule, priorities, or how David works.
David's Claw
Projects
Agentic systems, agent memory, harness engineering, and LLM orchestration—built for scale.
2024–2025
Multi-Modal Agentic Pipeline
Agentic AI for large-scale video processing—metadata extraction at fine granularity. Custom prompt caching, LLM routing, and inference optimization for order-of-magnitude speedups.
2024–2025
Explainable Screening Agents
Orchestrated agent layers for high-throughput, explainable screening workflows. Fast inference, compliance-first design, and elimination of manual bottlenecks.
2024–2025
LLM Harness & Document Pipelines
Harness engineering with structured retrieval, vector DB, and vLLM batching. Parsed domain content, optimized inference, and broad adoption.
2024
Distributed AI Assistant
High-throughput inference backend—Rust/C++ microservices, IaC, CI/CD, observability. RLHF and re-ranker for retrieval alignment. Horizontally scalable.
Research · 2024
Agent Memory & Retrieval
Published: scalable agent memory for code-generation agents. Parsed, classified, vectorized conversation and code logs for semantic retrieval—solving context fragmentation in long sessions.
Education
Carnegie Mellon University
Master of Information Systems Management
Deep Learning, NLP, Machine Learning, Distributed Systems, Relational Database
UCLA
B.A. Psychology (Computing) · Minor: Data Science & Statistics
OOP, CS, Computational Optimization, Statistics & Probability