Product Catalog Search
Agentic RAG over structured product records. Hybrid retrieval (BM25 + vector), faceted filtering, product comparison, and project cost estimation. Same pipeline, swappable retrieval engines so you can see the tradeoffs side by side.
Try any engine
HR Policy Knowledge Base
Document-level RAG with semantic chunking, parent-document retrieval, and multi-country policy filtering. Same architecture, different data model.
Featured Queries
Click any query to run it through the agentic RAG pipeline against the live catalog.
Why I Built This
B2B industrial distributors have a search problem that keyword search alone can’t solve. Buyers ask in plain English — “what do I need for a Cat6A drop end-to-end?” — but products live in dense technical spec sheets like Belden 7919A, 23 AWG F/UTP, CMR-rated. Project-style queries that span electrical, datacomm, security, and utility products widen the gap further.
The approach here is hybrid retrieval — BM25 plus dense vectors fused via Reciprocal Rank Fusion — wrapped in an agentic layer that uses Claude Sonnet 4 to plan multi-step retrieval for comparison, advisory, and project-bill-of-materials queries. A cheaper Claude Haiku 4.5 router handles classification and off-topic gating before the expensive model runs at all.
The same pipeline runs against Azure AI Search, Solr, Elasticsearch, and Google Vertex AI in parallel, so the architectural tradeoffs are visible side by side. On the 15-query golden set: 100% routing accuracy, NDCG@10 of 0.871 on Azure hybrid+semantic, and ~120 ms median latency. The full numbers — including an ablation across four embedding models and a RAGAS judge run by Claude Haiku — are on the evaluation page →
What I Bring
I combine deep AI/ML technical fluency — vector search, LLM-enabled relevance, RAG architectures — with strong customer empathy and UX instincts to deliver intent-aware, scalable search experiences. Proven record of defining product vision and OKRs, leading cross-functional teams, managing product P&L, and communicating strategy to senior leadership through data-driven storytelling.
Search & AI
Product Leadership
Professional Experience
JP Morgan Chase
Executive Director – Product Management Leader (Search & Discovery)
November 2010 – May 2025
Chicago, IL
- Owned end-to-end product vision, roadmap, and OKRs for a multi-application search platform serving 300,000+ users
- Championed migration to Amazon Kendra and directed a prototype RAG solution leveraging LLM-enabled relevance and vector search
- Led relevance tuning strategies delivering 15% increase in top-result relevance and 20% lift in platform engagement
- Built A/B testing framework using Adobe Analytics to validate ranking configurations and relevance models
- Managed global distributed agile team across engineering, design, AI/ML, analytics, and business stakeholders
Accenture
Application Architect – Digital Platforms & Search Strategy
February 2000 – November 2010
Chicago, IL
- Led technical strategy and delivery for enterprise search platform, designing search engine architecture and ranking logic
- Directed CMS re-platforming and multi-site digital transformation initiatives for B2B enterprise clients
- Supervised distributed, cross-geography development teams for mission-critical digital assets
Discover Financial Services
Senior Programmer | Analyst
July 1998 – February 2000
Riverwoods, IL
Education
Bachelor of Science (B.S.)
Management Information Systems
University of Illinois at Urbana-Champaign
Certifications
AWS Certified AI Practitioner
AWS Certified Cloud Practitioner