Engagements · Priced up front

Production readiness for AI systems.

Most AI projects do not fail at the demo. They fail in the gap between a prototype that impresses a room and a system that holds up under real traffic, real budgets and real adversarial users. That gap is the work I do.

Book a call nitish@nits.dev
Three ways to work

Scope and price, stated.

Ranges, not quotes — the number lands inside the range once I know the shape of the system. If none of these fit, say so and I will tell you honestly whether I am the right person.

01

Production-readiness review

1–2 weeks·$4,000–8,000

Architecture, retrieval quality, cost model, eval gaps and a prioritised remediation plan. Delivered as a written report.

  • Architecture and retrieval review against the 30-point checklist
  • Cost and latency model with per-request unit economics
  • Evaluation gap analysis — what you cannot currently detect
  • Prioritised remediation plan your team can execute without me
02

Scoped build

4–10 weeks·$12,000–35,000

I build the subsystem, harden it, and hand it over to your team with the tests and documentation that make it theirs.

  • Retrieval pipeline built and measured, not assembled and hoped for
  • Evaluation harness wired into CI with a pass threshold
  • Observability — tracing, quality alerting, cost attribution
  • Production hardening and a documented handover to your team
03

Fractional AI systems lead

Ongoing·$8,000–14,000 / month

Senior depth on one hard subsystem, without the hire. For teams who need the judgement more often than they need the headcount.

  • Technical direction on one system, held over months not sprints
  • Design review, code review and eval review on a standing cadence
  • Escalation path when something breaks and nobody knows why
  • Hiring support — I would rather make myself unnecessary
How I work

I work directly with the engineers building the thing, not through a layer of account management. Scope is agreed in writing before anything starts, and I would rather tell you a project is not worth doing than bill you for finding out slowly. Everything I build gets handed over with the tests and documentation that let your team own it — the goal is that you stop needing me.

A good fit
  • You have an AI feature in production, or close to it, and something is wrong
  • You have engineers — I work with your team, not instead of it
  • The problem is retrieval quality, evaluation, cost, or reliability
  • You can make a technical decision without a six-week committee
Not a fit
  • You want a prototype to show investors next week
  • The budget is contingent on the outcome being what you already believe
  • You need a full product team — this is one specialist, deliberately
  • The requirement is a specific vendor rather than a specific outcome
Next step

Tell me what is breaking.

Thirty minutes, no pitch. If I am not the right person you will leave the call knowing who is.

Book a call Read the checklist first