Let's talk AI products.
I'm always interested in conversations about AI product strategy — what's working, what isn't, and what we're all getting wrong. Hit me up on LinkedIn or drop a line.
AI Product Manager
Currently building personal AI infrastructure and shipping experiments in public. I think most AI product failures are design problems that got mistaken for engineering problems.
02 — Proof
⚙ AI Automation
Gmail → AI → Notion Pipeline
My inbox was a task graveyard — emails with action items I'd open, process mentally, and forget. I built a pipeline that watches Gmail, runs every message through GPT-4 to extract a structured task (title, priority, inferred due date), and upserts it into a Notion database with a threadId idempotency key so nothing duplicates on re-runs.
The interesting decisions weren't the code — they were the signal/noise filter design and figuring out what "a task" actually means to an LLM versus a human. Currently paused while I rethink the capture layer.
🔬 Local AI Research
Notion RAG — Local-First Knowledge Retrieval
Most AI search tools trade your data for convenience. I wanted neither the tradeoff nor the bill — so I built a local RAG system to query my entire Notion workspace using natural language. Stack: Ollama for local LLMs and embeddings, Chroma as the vector store, Streamlit for the interface.
The interesting part wasn't the build — it was the eval framework I built alongside it. Using a golden question set and Precision@5, I tested chunking strategies and overlap configs until retrieval quality was measurable, not just vibes.
03 — Point of View
Most AI features fail because nobody defined what good output looks like.
That's a product spec problem, not a model problem. The PM's job is to write the eval before the engineer writes the prompt.
The best AI PMs I know are the ones who've broken their own automations.
Shipping a personal AI tool teaches you more about failure modes, latency tolerance, and user trust than any case study.
AI is making product strategy harder, not easier.
When any feature is buildable in a week, the constraint shifts entirely to judgment: what should we build? That's a harder problem with more surface area for bad decisions.
I'm always interested in conversations about AI product strategy — what's working, what isn't, and what we're all getting wrong. Hit me up on LinkedIn or drop a line.