A selection of projects where I owned the outcome, from initial spec to final deployment.
A lightweight wrapper around the OpenAI API to standardize meeting notes and executive briefs. My first foray into API handling and system prompting.
Built a conversational AI assistant that answers questions about support ticket data using Retrieval-Augmented Generation. Users ask natural language questions like 'What are my most common problems?' and receive synthesized answers with source citations. Demonstrates complete RAG implementation plus comprehensive PM artifacts (UX design, evaluation methodology, cost analysis).
A React + Node.js application that bridges the gap between raw API calls and safe enterprise usage. Features a 'Template Library' for standardized tasks (Support, Summarization) and an 'Ethics Checkpoint' that requires human validation before sending data to the LLM.
A specialized model pipeline that ingests raw support tickets and outputs structured JSON categorization, optimizing for unit economics over raw intelligence.
An AI-powered analytics platform that automatically discovers patterns in unstructured customer feedback using unsupervised learning. Goes beyond typical sentiment analysis to cluster tickets, identify systemic issues, and generate executive summaries—demonstrating enterprise-grade data governance (PII protection, cost controls, audit trails) alongside ML implementation.
An agentic workflow tool that transforms messy CSV backlogs into prioritized, INVEST-ready user stories. Features a 'Reflexion Loop' for self-correction and RICE scoring for strategy.
An autonomous AI agent that eliminates the 'training data cutoff' problem by intelligently routing between internal knowledge and live web search. The system decides which data source to query based on intent—no manual intervention required. Demonstrates advanced agentic workflows with production-validated cost telemetry ($0.26/query) that enabled downstream ROI modeling.
A financial modeling tool that translates AI system capabilities into executive-level business justification. Uses real production telemetry from Block 6 ($0.26/query) to generate 3-year cumulative ROI projections with cited industry benchmarks. Demonstrates critical PM skill: communicating technical value in business language that secures budget approvals.