ESARC
Contact us

Services - Three ways we engage.

Every engagement is led by a principal engineer and backed by our in-house AI agent fleet. The shape changes with the problem, from a short diagnostic to a focused build to an embedded AI team.

2 weeks · fixed scope

AI Diagnostic Sprint

A short, written assessment of an AI surface or roadmap. We read the code, talk to the team, ship a prioritised plan with an exec readout.

Start this engagement
Who it’s for
You’re mid-build, or about to start, and you want a senior outside read before you commit a quarter of engineering time.
Pick this when
You need clarity before you commit. Or your board wants a sober second opinion on the AI roadmap before they greenlight the spend.
Typical scope
Architecture, model choice, eval coverage, latency, cost, security, team gaps. We don’t write production code on this engagement. We write down what to build next.
What you get
  • A written audit, 20 to 40 pages, not a slide deck
  • A prioritised 90-day roadmap with effort estimates
  • A 60-minute exec readout, recorded
  • A draft eval rubric for the top two opportunities
  • A short list of what to stop doing
Related work
McGraw Hill

6 weeks · fixed scope

Build Sprint

One AI surface, shipped. LLM feature, voice agent, eval harness, RAG system, or agentic backend. Daily PRs, eval gate, staged rollout.

Start this engagement
Who it’s for
You have a concrete AI feature in mind and you want it in production this quarter, not next year. Scope is clear enough to lock.
Pick this when
Scope is clear. You’ve picked the surface. You want it shipped and instrumented in six weeks, not adopted into a multi-quarter program.
Typical scope
Single surface, single outcome. Voice agent, RAG search, LLM-assisted workflow, model eval harness, agentic backend. We turn down scope creep, on purpose.
What you get
  • A principal in your repo from day one
  • Daily PRs against a feature branch with eval gates
  • A working build behind a feature flag by week four
  • Staged rollout (1%, 10%, 50%, 100%) by week six
  • Eval suite, dashboards, runbook checked into your repo
  • 30 days of post-engagement on-call
Related work
Stuf — Sidney Voice AIScrubs Co-Pilot

3-6 month embedded team

Embedded AI Team

Principal plus agent fleet inside your repo and your Slack. Output on the order of a team of senior engineers, with one accountable owner.

Start this engagement
Who it’s for
You have an AI roadmap, not a single feature. Multiple surfaces, ongoing work, no senior engineer to own it, and you don’t want to hire one in a panic.
Pick this when
AI is a roadmap, not a feature. You need senior accountability on the work week in and week out, and you want output that scales past a single partner’s hours.
Typical scope
Owning a real product surface across months. Voice, agents, RAG, evals, infra. You set priority, we set sequence. Output is measured in PRs, evals, and dashboards, not status reports.
What you get
  • A dedicated principal in your Slack and standups
  • An agent fleet running on your repo with your guardrails
  • Daily PRs, weekly written review, monthly readout
  • A rolling technical roadmap, updated monthly
  • Quarterly exec reviews
  • Clean written handoff at the end of the engagement
Related work
MyMethodSpringhouseMeta Superintelligence Labs

Common questions

Stuff buyers ask before they sign.

Who owns the IP?

You do. Standard work-for-hire, signed before the first PR. Code, prompts, evals, weights you fine-tune on your data, all yours.

Where do the agents run?

Two options. Default: in our cloud, against your repo via a scoped GitHub app and a tight allowlist. Enterprise: inside your VPC, on your provider, with your audit logs. We do not store your code at rest beyond the engagement.

What models do you use?

Whatever fits the job. Claude Opus and Sonnet for code and reasoning, GPT-4o and 4.1 for general work, Llama for self-hosted, ElevenLabs and Vapi for voice. We pick per task and tell you what we picked.

How does the handoff work?

Last phase of every engagement is handoff. Runbook, evals checked into your repo, on-call playbook, a recorded walkthrough. We stay on call for 30 days after we leave.

What about security and compliance?

We sign MSAs, NDAs, and BAAs where applicable. SOC 2 documentation is in progress. For regulated work we run agents inside your environment, never ours. We’ve shipped against EPIC, HIPAA-scoped data, and AWS production at Amazon-scale.

What time zone do you work in?

EST and PT, from Canada. That covers the entire North American working day on a single contract. Most ongoing work is async-first, so timezone usually matters less than people think.

How many engagements do you run at once?

Two or three at a time, never more. If you have a real engagement, you get real attention.

What if it doesn’t work out?

Diagnostic Sprints are short, so the answer is simply done. Build Sprints have a checkpoint at week two where either side can walk. Embedded engagements have 30-day notice. We’d rather you leave clean than stay unhappy.

Tell us what you’re trying to ship.

A principal engineer reads every inbound. We reply same day on weekdays, with an honest read of whether we’re the right team for the work.

Start a conversationOr book a 30-min intro →