How to choose a vector database
Talk abstract. Reading the vector pricing calculators, linking features to outcomes, and what else decides it anyway — existing search systems, AWS credits, and how fast your data grows.
adam hevenor
One guy,
many agents.
AI, search, and agentic engineering consulting.
I help CTOs turn their engineering org into a technical staff that directs agents toward outcomes. Recent work: 500M clinical-note chunks, in production.
consulting case study
A search api that became an agentic platform
Trio Health builds custom datasets for highly specific rare diseases. How we designed an agentic search system capable of extracting meaningful insight from billions of clinical patient notes.
AWS EKS + KEDATurbopufferBM25 + semantic hybridGPU embedding pipeline
no decks
Every proposal ships an agent.
I don't pitch with a slide deck. When we talk about working together, I send a proposal you can actually interrogate — an agent that's read your context and the work I'm proposing, and answers your team in my voice. Hire me and it stays: every engagement keeps its agent as the leave-behind, fed by everything we produce.
- Before you commit. Kick the tires on the plan, not a PDF.
- In my voice, grounded. Not a generic bot — me, citing the actual material.
- Yours to keep. Engagements keep the agent, fed by the whole project.
training
Make Your Product Agent-Ready.
Cohort courses and team trainings for leaders and builders who need practical AI engineering fluency. The flagship: Make Your Product Agent-Ready, a 3-week cohort. A Build an Agent course is in development for Q4.
June 25 — free lightning lesson on Maven: Three Patterns for Better Agentic Search -> co-hosted with Doug Turnbull & Trey Grainger (AI-Powered Search)
services
view services -> Workshops A hands-on half day with your team: operating an agent factory, evals from zero, KPIs that drive behavior, or one shared vocabulary across your leadership. Your team leaves with something working. Research Plan A hosted plan for the question you're actually trying to answer — what to learn, from whom, in what order, and what would change your mind. Your team marks progress against it and interrogates it with an agent. Research Program The plan, and I run it. Six interviews with people who can answer the question, then the insight analysis: what the market said, where it contradicts your roadmap, and what to do about that. Search & Retrieval POV A proof of value on your own corpus. Evals first, so "better" is a number you can defend, then hybrid, reranking, chunking, and the vector database call, argued from your data. You keep the eval set and the system. Fractional CTO I run a factory on your codebase. Agents do the volume, I own the review gate. Around that: architecture, vendor selection, hiring judgment, and delivery rhythm. Hourly Data engineering, ML, and strategy work billed by the hour — for a bounded problem, a second opinion, or ad-hoc help when you don't need a standing engagement.
open source & r&d
view r&d ->The practice funds focused search-infrastructure R&D, shipped under the hev brand.
recent writing
view all ->Dope tech: TopK
First entry in a series on the search tech I actually reach for. TopK gets the 2026 architecture right — multi-vector late interaction, object-store-native plus NVMe, LSN read-after-write consistency, and Postgres-compatible SQL. Not enterprise-ready yet, and I say where.
A new role for SWE, managing outcomes not outputs
I haven't reviewed a line of code in over a decade — not because I can't read it, but because I learned in 2008 that trust comes from outcomes, not from meddling in output. That's the same lesson every engineer handing work to an agent is about to learn.
Competing as a privateer in the AI race
Racing is the ultimate expression of engineering, and the privateers — people building their own agent harnesses and competing at the frontier — are where the real innovation in AI is happening in 2026. A field guide, from F1 and downhill MTB to Geoffrey Huntley, OpenClaw, and Pi.