Skip to content
rene persau
menuclose
services

Three ways to work together.

Fixed scope where the outcome is clear, a retainer where it isn't yet. Every engagement is scoped on a call first, so you know the deliverables and the timeline before anything is signed.

how pricing works
fixed scope · quoted per engagement
retainer · monthly, capped days
audit · two weeks, flat fee
no hourly billing
01 · fixed scope

AI agent build

4–8 weeks

For teams whose agent works in a demo and fails on real requests. I design the retrieval, planning and tool layers it needs, then build it in your repo alongside your engineers.

  • Retrieval & indexing strategy
  • Hybrid retrieval + reranking
  • Planning & routing
  • Multi-agent orchestration
  • Tool & API integrations
  • Traceable answers & actions
you end up with

A running system in your repo, an architecture note explaining every decision, and a golden set that tells you when it breaks.

scope this →
02 · fixed scope

Evaluation & guardrails

3–5 weeks

For systems already in production with no way to tell whether a change made them better. I build the measurement layer: a golden set from your real traffic, evals in CI, tracing you can read, and hard limits on what the system may do.

  • Golden set from real traffic
  • Offline & online eval suites
  • Regression gates in CI
  • Tracing & observability
  • Policy & safety guardrails
  • Human escalation paths
you end up with

A release gate your team owns. Prompt and model changes stop being a gamble, and a regression blocks a deploy instead of reaching a customer.

scope this →
03 · retainer

AI development partner

monthly · capped days

For teams building AI as a core part of the product who want a senior engineer in the room continuously. Hands-on delivery, architecture review, and the unglamorous work of making a system observable and operable.

  • Hands-on delivery with your team
  • Architecture & PR review
  • Roadmap & build-vs-buy calls
  • Hiring input & tech screens
  • Incident support on AI paths
  • Team upskilling sessions
you end up with

A team that ships AI features without me. Most retainers get smaller over time, which is the point.

talk it through →
04 · process

How an engagement runs

  1. week 0

    Scope call

    One hour on your data, your users and the failure you're seeing. You leave with an opinion whether or not we work together.

  2. week 1

    Baseline first

    Golden set and measurement before any architecture change. Everything after is judged against it.

  3. weeks 2–6

    Build in your repo

    Weekly demos and PRs your engineers review. No black boxes, no parallel codebase to inherit.

  4. close

    Handover

    Runbook, eval suite, architecture notes, and a working session with the engineers who take it forward.

Not sure which one fits?

Send a few lines about what's breaking. If a two-week audit is the honest answer, I'll say that instead of selling you a build.

Describe your problem →