I'm a go-to-market leader running HPE's private cloud portfolio. I'm also completely AI-pilled and I love building.

An enterprise-grade RAG pipeline for question answering over enterprise PDFs. Keeping retrieval quality high on real, messy documents is the actual problem, so this is built like infrastructure: Qdrant vector search, Celery workers for async ingestion, and an LLM evaluation framework that scores retrieval quality instead of guessing at it.

The eval framework is the part I care about most. If you can't measure retrieval quality, you can't improve it.

AI agents break quietly. A prompt tweak or model update changes behavior and nobody notices until it's in production. AgentCI is a pytest-style regression testing framework for agents: it records traces, detects behavioral and cost regressions, and runs in CI/CD like any other test suite. No LLM-as-judge required, so results are deterministic and cheap.

Open source, Apache 2.0, published on PyPI as ciagent.

AI customer service for Shopify merchants. Most tools in this space charge per conversation, which punishes merchants for growing. Approved and live on the Shopify App Store.

I built the GTM too: an automated Reddit pipeline that surfaces merchants asking for help, and a free store scanner that grades how well a storefront answers customer questions. The scanner has graded 100+ stores so far.

Document redaction for lawyers where nothing ever leaves the machine. Cloud redaction tools ask firms to upload privileged documents to someone else's server, which is exactly the problem. RedactLocal runs entirely on-device.

Under the hood it's a four-layer pipeline: regex and checksum matching, then GLiNER zero-shot NER, then optional Gemma 4 validation, then mandatory human review. Policy packs cover FRCP 5.2, HIPAA, GLBA, and immigration documents.