Expert Council
LiveMany viewpoints collapsing into clarity.
Expert Council assembles a panel of AI personas to debate your question in real-time, preserving disagreement and trade-offs instead of flattening them into a single answer.
The work speaks now.
I'm Rob Spence — Solutions Architect by day, AI tool builder by compulsion. The projects on this site are things I built because I needed them to exist: real tools I use, not demos I forget about. Most of them run on Claude. All of them are fully shipped. This portfolio is my proof of work.
Many viewpoints collapsing into clarity.
Expert Council assembles a panel of AI personas to debate your question in real-time, preserving disagreement and trade-offs instead of flattening them into a single answer.
AI-powered technical interviews with lens-based evaluation and DSPy-optimized scoring. Not a quiz — a structured conversation that reasons about candidates.
The laziest AI ever built. On purpose.
Personality-driven agents that break your app the way real users do.
I built this to stress-test my own AI tools before demos. It runs a structured evaluation suite against any AI app — response quality, edge case handling, hallucination signals. I use it before every OutSystems SA engagement where I'm showing AI tooling.
Autonomous equity analyst. S&P 500 stocks scored through 8 investing experts (Buffett, Lynch, Graham, Damodaran, Munger, Greenblatt, Marks, Smith) across four pillars — business quality, competitive moat, financial health, and governance — using SEC 10-K filings, yfinance metrics, and Claude.
An AI heist planner cases your job and delivers the personalized takeover blueprint.
The graph-based memory layer that prevents Groundhog Day in AI.
Upload a schema, optimize it with DSPy/MIPROv2, deploy as a REST API — 10–30x cheaper than raw GPT-4o.
Satellite imagery + ML segmentation = instant lawn care quotes from space.
API-first spreadsheet analysis. The most shipped product in the catalog.
APIs and infrastructure — deployed services, not demos.
Decision Archaeology: reconstructs what was decided, what was explicitly rejected, and why — from public GitHub PR history.
My personal AI operating system — running on my Mac, live in my Slack, in daily use for the past four months. It spawns workers, files research to Obsidian, reads my calendar and email, and surfaces what matters. Not a prototype. Not a concept. This one I actually depend on.
Open-source tools built for real workflows — not demos, not prototypes.
Research and investigations — not production tools, but real findings.