Work with Andrew
03 / 13
Production AI systems
Senior/staff work on AI systems that must hold up in production: speech and audio pipelines, RAG, LLM orchestration with compliance controls. Ran Abstract's AI practice as Director of Engineering through 2026: FINCA's six-language audio pipeline, ALL-SAFE's prize-winning training assessment, Saatva's customer intelligence.
Credentials
- FINCA International: thousands of hours of field interviews across 4 countries and 6 languages, searchable in seconds
- ALL-SAFE: agentic trainee evaluation on the $700K grand-prize platform, deployed across 3 countries
- Saatva: customer intelligence over calls, chat, reviews, and support tickets
Why these lanes belong together
The same pattern keeps returning in my work: someone has a system nobody wants to take seriously, and it turns out to be load-bearing. Vacation-home water heaters; a mold inspector's leave-behind sensors; forty federal datasets; surgical training in hospitals with bad connectivity. Every home in a 320-property rental fleet got a digital twin. The mold sensor went through contract manufacturing. The datasets became APIs that AI agents query directly. The surgical platform's prize was $700,000. It runs in Ethiopia. That is the useful part.
Good fit
- Ambiguous zero-to-one systems where the technical direction and the implementation are the same job.
- Data-heavy AI products that must operate reliably in production, under real compliance constraints.
- Problems that cross boundaries: software to hardware, engineering to institutions, prototype to operations.
- Small teams that need a lead who ships alongside them.
Not a fit
- Pure research roles judged on publications. The AI-mechanism work is applied engineering, not an academic track.
- Roles that are exclusively people management with no hands-on technical scope.
- Teams seeking a narrow specialist in a single framework. The value here concentrates where problems cross domains.
- On-site-only roles outside Austin, TX.
A useful first email says what you're building, where it hurts, and what form of help you think you need. Specifics get the fastest reply. hello@andrewfribush.com. The full record is at the engagement ledger, the short version at the résumé. Machine-readable: /career/brief.json · /availability.json.