How we work
Engineers who deploy with you.
Forward Deployed Engineering
A Headrun pod embeds in your team, takes one workflow to production and stays accountable until it holds.
01 / Definition
A forward-deployed engineer is a senior engineer who works inside your operation, in your repo, your data and your standups, and answers for a system reaching production. Not for hours billed.
02 / Three models
Three ways to buy AI engineering.
| Traditional consultancy | Staffing agency | Headrun FDE pod | |
|---|---|---|---|
| What you buy | Advice and a roadmap | Hours | A system in production |
| Who shows up | A pitch team, then a delivery team | The people available that month | A lead and engineers who stay for the whole engagement |
| Where they work | Their office and your boardroom | Your ticket queue | Your repo, your tools, your standups |
| First deliverable | A slide deck | A filled seat | A working slice of the system, with evals |
| Who owns the architecture | They recommend it, you build it | You do | The pod does, with your team |
| Accountable for | The recommendation | Attendance | The metric agreed at the start |
| When scope changes | A change request | A new requisition | The pod re-plans with you that week |
| What you keep | A document | Whatever was written while they were there | Code, evals, runbooks and a team that knows the system |
03 / Anatomy of a pod
Small, senior, complete.
One lead who owns the architecture and the relationship. One or two AI engineers who build. Specialists in data, evals and interface who join when the work needs them.
Always in the pod
Lead
Owns the architecture, the plan and the relationship. Your single point of accountability.
Always in the pod
AI engineers
One or two engineers who build in your repo and ship through your release process.
Join on demand
Specialists
Data, evals and interface engineers who attach when the work needs them and leave when it does not.
04 / The engagement
Four phases. A gate at the end of each.
Each gate is a named state both sides can check. At every one you can stop with something usable, or continue.
Phase 1
Scope
One call, one page. We name the workflow, the metric and the first thing we would ship.
Gate 1 · Metric agreed
Phase 2
Embed
Engineers join your standups, your repo and your chat. They map the workflow as it really runs, not as it is documented.
Gate 2 · Workflow mapped
Phase 3
Ship
A working system in production on a narrow slice of real work, with evals from the first commit.
Gate 3 · Live on one slice
Phase 4
Harden
We widen the scope, tighten the guardrails and hand your team the runbook. Or you keep us on.
Gate 4 · Runbook handed over
What each gate means
| Gate | Passed when | You leave with |
|---|---|---|
| 1. Metric agreed | The workflow, the success metric and the first release are written in the scoping note. | A one-page plan you could hand to any team. |
| 2. Workflow mapped | The pod has access and has documented the workflow as it really runs. | An accurate map of the process and where it breaks. |
| 3. Live on one slice | The system handles a defined slice of real work in production, with evals passing. | A working system and its eval suite. |
| 4. Runbook handed over | Your team has taken the system through a release and an incident without the pod. | Code, evals and runbooks, plus the option to extend. |
05 / What we need from you
Four things, in week one.
An engagement moves at the speed of access. These are the only things we ask for.
- A named owner
- Access to the systems in week one
- One workflow
- One metric
06 / Inside your walls
Your cloud, your repo, your keys.
We work under your access controls and your NDA. The code is yours from the first commit.
Access
Under your controls.
The pod works with the accounts, roles and review rules you grant, and nothing wider.
Data
In your environment.
Systems are built and deployed in your cloud, against your data, where your policies already apply.
Ownership
Yours from commit one.
Code, evals and runbooks live in your repo throughout the engagement, not handed over at the end.
07 / What you keep
Assets, not dependency.
01
The code in your repo.
02
The eval suite.
03
The runbooks.
04
A team that has worked alongside ours.
08 / Questions
Before you ask.
Is this staff augmentation?
No. You are not renting people to manage. The pod brings its own lead, its own way of working and a commitment to an outcome we both wrote down.
Do we need clean data first?
No. Finding and fixing the ground truth is usually the first week’s work.
Which models and clouds do you use?
Whatever fits the workload and your constraints. We route between models on cost, latency and accuracy, and we deploy in your environment.
Where do your engineers sit?
Our engineers are based in Bengaluru. Overlap hours and time on site are agreed in the scoping note, before the engagement starts.
What happens when the engagement ends?
Your team runs the system with the runbooks and the evals. Or the pod stays and takes the next workflow.
How is it priced?
Scope and price are written into the scoping note, so you see both before any work starts.
09 / Start
Put a pod on your hardest workflow.
Tell us which workflow. An engineer will tell you what a pod would ship first and how we would measure it.
You will talk to an engineer