Rijad Tinjak.
Founder & chief architect, starmo
Rijad Tinjak is the founder and chief architect of starmo, the engineering studio he runs with Mesa. They build AI systems that survive production: retrieval assistants with cited answers, agents on deterministic guardrails, and monitoring systems that run daily. Five shipped systems, two of them live SaaS products starmo operates, all engineered with the test coverage, cost tracing, and evaluation discipline documented on this site.
- Production AI systems shipped
- 5
- Automated tests across them
- 1,609
- Live products designed, built and operated
- 2
- Per-query cost of production retrieval, measured
- $0.006–0.02
How we work
End-to-end, hands-on, and measured. Every system on the work page was built full-stack: ingestion pipelines, retrieval cores, backends, interfaces, tests, and deployment infrastructure, with per-query cost tracing and evaluation sets treated as part of the architecture, not afterthoughts. The engineering positions we defend: hybrid retrieval over naive embedding search, deterministic guardrails over prompt-based safety, golden eval sets over vibes, pgvector over premature vector-database complexity, and no framework where two hundred lines of owned code will do.
Selected systems
RegNotify
starmo product · Live
Fleet-aware regulatory monitoring SaaS for Part-145 MROs, CAMOs and smaller airlines, live and shipping weekly.
Read the case studyDocsAI
Leav Aviation · Production
RAG and agentic assistant over aviation manuals and live operations data. Every answer cited to its source page.
Read the case studynowreply
starmo product · Live
AI front desk for appointment businesses: answers every inquiry from the business's own prices, policies and calendar, and books the appointment.
Read the case study
All the systems, with verified numbers: the work page. First-party measurements: Production AI Benchmark 2026.
Talk to us directly.
The people on the call are the people who build the system.