Featured Work

Projects and initiatives from the past several years.

AI Teammates

AI Product

Led engineering for Asana's AI agent product – three teams – from private beta through general availability. The work centered on narrowing launch scope, improving reliability, and shipping GA on time.

  • GA launch delivery
  • Reliability engineering (86% → 99%)
  • Launch scope focus
  • Retention above launch targets
Agentic AI Product Leadership Reliability GA Launch

AI Vendor Partnerships

Strategic Partnerships

Led Asana's vendor relationships with Anthropic, OpenAI, and Mistral – early access programs that enabled key product capabilities, one of the first production MCP servers, and an appearance in Anthropic's Claude 4 launch video.

  • Anthropic partnership (Claude 4 launch feature)
  • Production MCP server
  • OpenAI partnership
  • Mistral partnership
Partnerships Anthropic OpenAI MCP

RAG Platform Development

AI Infrastructure

Led the team behind Asana's RAG platform, a shared retrieval layer used across Asana's AI features. It grew from a single search feature to a shared framework handling 320k weekly requests.

  • Advanced vector search capabilities
  • Context-aware document retrieval
  • Cross-application integration framework
  • Cut p95 request latency of onboarded features by >50%
RAG LLMs Vector DB Performance

AI Evaluation System

Quality Assurance

Built Asana's large language model testing and evaluation program – a framework for pre-release model testing and AI feature quality.

  • Automated evaluation suite
  • Quality metrics and benchmarking
  • Pre-release model testing
  • Strategic partnerships with Anthropic and OpenAI
Evaluation LLMs QA Benchmarking

Agentic Context Engineering

Performance Optimization

Led the program developing context engineering techniques for agentic AI systems, improving performance and quality while cutting operational costs. The work led to a patent application.

  • 24% improved time to first token (TTFT)
  • Enhanced evaluation pass rate into the high-90s
  • 35% LLM budget reduction
  • Context optimization algorithms
Agentic AI Performance Cost Optimization Context Engineering

ML Recommendation System

Machine Learning

Implemented a recommendation system for Shift's car marketplace that measurably moved primary conversion metrics.

  • 25%+ increase in primary conversion rate
  • Personalized recommendation algorithms
  • User behavior analysis
  • A/B testing framework
ML Recommendations Conversion Product

Bit Vector Algorithm

Performance Engineering

A custom bit vector-based index for critical filtering and sorting operations, cutting processing time from about 7 seconds to 7 milliseconds.

  • Reduced processing time from 7s to 7ms
  • Optimized memory usage
  • Scaled to handle large datasets
  • Improved user experience
Algorithms Performance Optimization Scalability

Third-Party Integrations Framework

System Architecture

Extended Asana's RAG platform into third-party data sources, adding connectors for external tools alongside Asana-native content.

  • Google Drive integration
  • OneDrive connectivity
  • Amazon Q integration
Integrations API Design Data Connectors Architecture