# Josh Solis

> Josh Solis is an AI engineer and consultant in the Dallas-Fort Worth area, working remotely and traveling to clients. Josh Solis builds AI systems that ship: computer vision, AI agent platforms, and the LLM eval harnesses that prove they work.

Josh Solis worked for years in sales, account management and business ownership. Josh Solis scopes work like a consultant and builds like an engineer. Background: Rice University, B.A. Cognitive Science, 2017; a nine-month AI and ML bootcamp, 2019-2020; about five years in business. 5,300+ commits in ten months across all projects, including 4 featured systems. Client work uses codenames. Not the chef, the MMA fighter or the musician with the same name.

## Case studies

- [Blueprint Vision: Computer vision for construction takeoff and HVAC estimating](https://joshdsolis.com/work/blueprint-vision/): Computer vision for construction takeoff. Models find HVAC symbols on drawings, and each human fix becomes new training data. Best detector: mAP50 0.755; Automated tests: ~1,100; Commits: ~1,900.
- [Collectible ID: Near-duplicate image identification and pricing at catalog scale](https://joshdsolis.com/work/collectible-id/): A phone app finds the exact printing of a collectible in a 151,260-item catalog. Then it attaches a market price. Catalog items: 151,260; Review-queue query: 90 ms to 3 ms; Surfaces: API · Web · Mobile.
- [Agent Systems: AI agent platforms: escrow, spend limits and a hardened agent fleet](https://joshdsolis.com/work/agent-systems/): AI agents compete, get paid, and run with small, revocable authority. 28 agent roles run on a locked-down server. Agent roles: 28; Open-source repos: 2; Exposed SSH ports: 0.
- [Harness Engineering: LLM eval harnesses and AI agent evaluation with pass / fail / error verdicts](https://joshdsolis.com/work/harness-engineering/): LLM eval harnesses and test gates that check every change to code, models and data pipelines. Harness types: 6; Automated tests (one product): ~1,100; Commits / 10 mo: 5,300+.

## Consulting

- [Work with Josh Solis](https://joshdsolis.com/hire/): fixed-scope AI consulting in the Dallas-Fort Worth area, working remotely and traveling to clients. Engagements start with a 30-minute intro call; quotes follow the call. Josh Solis replies within two working days.
- AI Feature Audit (3 days): Teams with an AI feature that feels unreliable. A written review of the feature, its data and its tests, and a fix plan in priority order.
- LLM & Agent Eval Harness Setup (2-3 weeks): Teams that ship LLM or AI agent changes without proof. An eval corpus, a scoring function, a CI gate with JSON verdicts, and a runbook.
- Computer Vision / AI Agent Prototype (3-4 weeks): Teams that want to test an AI idea on their own data. A working demo with metrics and a clear go / no-go note.
- AI Dev Workflow Setup (1-2 weeks): Small dev teams that want to use AI coding agents safely. Agent rules files, a worktree workflow, a verify command, and session hand-off tools.

## About

- [How Josh Solis works](https://joshdsolis.com/how-i-work/): a six-step loop (research, plan, build in parallel, gate, ship, learn) and a month-by-month record.
- [About Josh Solis](https://joshdsolis.com/about/): background, other areas of work, and FAQ.

## Field Notes (weekly learnings)

- [All Field Notes](https://joshdsolis.com/notes/): weekly notes from building with AI agents, evals and computer vision.
- [pass, fail, error: a three-valued verdict for AI agent loops](https://joshdsolis.com/notes/pass-fail-error/): Every check needs three outcomes: pass, fail and error. Why error must never count as either, and what that means for Karpathy-style agent loops.

## FAQ

- [What does Josh Solis build?](https://joshdsolis.com/about/#faq): Josh Solis builds AI systems that ship: computer vision models, AI agent platforms, and the LLM eval harnesses and test gates that prove they work.
- [Is Josh Solis available for AI consulting in Dallas-Fort Worth?](https://joshdsolis.com/about/#faq): Yes. Josh Solis takes fixed-scope AI consulting work. Josh Solis is based in the Dallas-Fort Worth area and can meet clients in person there.
- [Does Josh Solis work remotely?](https://joshdsolis.com/about/#faq): Yes. Josh Solis works remotely with teams anywhere and travels for on-site work, with a working demo each week.
- [What does an LLM eval harness setup include?](https://joshdsolis.com/about/#faq): An LLM eval harness setup by Josh Solis includes an eval corpus from the client's own data, a scoring function, a CI gate that returns pass, fail or error as JSON, and a runbook. It takes two to three weeks.
- [What is an AI Feature Audit?](https://joshdsolis.com/about/#faq): The AI Feature Audit is a three-day review of one AI feature, its data and its tests. The client gets a written report and a fix plan in priority order.
- [What is harness engineering?](https://joshdsolis.com/about/#faq): Harness engineering is building the checks around AI work. An LLM eval harness grades each change to code, models or data and returns pass, fail or error. Error means the check could not run, and it never counts as a pass.
- [What computer vision work has Josh Solis done?](https://joshdsolis.com/about/#faq): Josh Solis trained YOLO and RF-DETR detection models that find HVAC symbols on construction drawings for takeoff and HVAC estimating. The best detector reached mAP50 0.755.
- [How does an engagement start?](https://joshdsolis.com/about/#faq): An engagement with Josh Solis starts with a 30-minute intro call, then a written scope and a quote. Most engagements take one to four weeks.
- [Can client details stay private?](https://joshdsolis.com/about/#faq): Yes. Josh Solis uses codenames for all client work and can sign a mutual NDA before seeing client data.
- [How fast does Josh Solis reply?](https://joshdsolis.com/about/#faq): Josh Solis replies to new requests within two working days.

## Optional

- [Full text](https://joshdsolis.com/llms-full.txt): every case study in plain markdown.
- [GitHub](https://github.com/joshdsolis)
- Contact: joshdsolis@gmail.com