A small studio with opinions about AI.
We build AI applications for organizations that want results — not just a chatbot. Eight years of applied ML and three years of frontier-LLM production experience, in one team.
Our mission
"Diyaan" means light, illumination. Most enterprise data is dark — locked in contracts, tickets, transcripts, sensors. We build the systems that make that data legible, decisionable, and finally useful.
We are intentionally small. Every engagement is led by a partner-level engineer. We turn down work that is not a fit, and we will tell you when the AI answer is "do not build it."
jurisdiction: Kentucky, USA
founded: 2026
headquarters: Richmond, KY
verticals: retail, healthcare, financial services, SaaS
governance: NIST AI RMF, EU AI Act-aligned
infra: AWS / GCP / Azure / on-prem / air-gapped
models: frontier, open-weight, custom fine-tunes
How we work
Use case first, model second
The right model depends on the problem. We benchmark frontier, open, and small models per task — no vendor loyalty.
Evals are the product
Without measurement you have wishful thinking. Every engagement begins with the eval harness.
Cost-per-success is real
Frontier models are not always the answer. We optimize the path from input to outcome, not the marketing benchmark.
Production is the bar
Demos are easy. We measure ourselves on uptime, latency, accuracy, and ongoing cost in your environment.
Governance from day one
Eval harness, model card, red-team battery, audit trail. Built in, not bolted on.
Boring tech where possible
If a regression beats the LLM, we ship the regression. Sophistication is not a goal.