AI staff augmentation: dedicated AI developers
Engineers who have shipped RAG systems and agents to production, embedded full-time in your team, on a rolling monthly contract.
AI staff augmentation means adding AI engineers from an outside firm to your own team for as long as you need them. They work in your repos, your standups and your tools, under your technical lead. Unlike outsourcing a project, delivery ownership stays with you: you direct the work, and you own what they write.
Here, that is a full-time AI engineer inside your team, onboarded within 48 hours of signing, at $2,800 per engineer per month. Every line of code, every prompt and every eval belongs to you from the first commit. One month minimum, 30 days notice, no annual contract. Weighing the two models? Read staff augmentation versus project outsourcing.
Who this is for
An AI feature is on this quarter's roadmap and nobody in-house has shipped retrieval or agents to production before. Your engineers are good - they are just being asked to learn evaluation, chunking strategy, prompt regression and token economics on a deadline, and the first honest estimate they give you is a number with a very wide error bar.
Hiring solves it in three to six months. A dedicated engineer solves it this week, and leaves your team able to maintain what was built.
What "dedicated" means here
- 01Full-time on your team. Not shared across three accounts, not time-sliced, not rotated away when a bigger client appears.
- 02In your systems. Your Git host, your Jira or Linear, your Slack, your cloud account. We do not run a parallel project tracker and email you a status report.
- 03In your standups. The engineer attends your ceremonies at your times and is managed by your lead, not by an account manager in the middle.
- 04Named and reachable. You know who they are, you met them before they started, and you message them directly.
- 05Not a ticket queue. There is no SLA to hide behind and no work order to raise. They are on your team.
The engineers
Profiles are anonymised here because most of these engagements are under NDA. Before you sign you get the named CV, the public GitHub history where there is one, and a technical call with the specific engineer you would be getting.
Senior AI engineer, 8 years
Python, FastAPI, Qdrant, LangGraph, Postgres. Built the hybrid retrieval layer behind the Klebbix platform - 68% latency reduction across multi-tenant enterprise workloads.
AI engineer, 6 years
TypeScript, Node, OpenAI and Anthropic SDKs, Temporal, Redis. Shipped an agentic code-repair loop with an 87% first-pass fix rate, running in isolated Docker sandboxes.
ML engineer, 7 years
Python, PyTorch, Cohere Rerank, evaluation harnesses, Grafana. Owns the eval harness we ship with every build: regression suites, drift alerts and a human review queue.
Full-stack engineer, 5 years
Next.js, React, Postgres, Azure AD SSO, Docker. Built the multi-tenant admin and RBAC layer for a GDPR and ISO-27001 scoped deployment.
You can read two of these systems end to end: the Klebbix multi-tenant retrieval build and the self-healing engineering agent.
Roles and vetting
We place engineers in three roles. Most teams start with one and add a second role as the work grows.
- 01AI engineer. Retrieval and agents: hybrid retrieval, re-ranking, agent orchestration and evaluation harnesses, in Python or TypeScript.
- 02ML engineer. Evaluation and model quality: regression suites, re-ranking, drift alerts and human review queues.
- 03Full-stack engineer. AI inside an existing product: auth and permission boundaries, admin interfaces, multi-tenant isolation and rollback paths.
How engineers are vetted
- 01Employees, not resold contractors. Every engineer we place is a MacroCoderz employee, from the team that built the Klebbix retrieval platform and the self-healing engineering agent.
- 02Checked before client work. Background checks, to the extent local law permits, a signed confidentiality agreement, and security training at onboarding and every year.
- 03Vetted by you. The named CV, public GitHub history where there is one, a technical call with the specific engineer, and your own take-home or pair-programming exercise if you want one.
- 04Reviewed once they start. Code review and architecture input from our senior engineers throughout the engagement.
How onboarding works
- 01day 0
Brief and match
You describe the work and the stack. We come back with the engineer we would put on it, their CV, and a time for a technical call. If we do not have the right person on the bench we say so and give you a sourcing date.
- 02day 0-1
Technical call
Your lead talks to the engineer. Run whatever exercise you normally run. If you say no, we propose someone else or we stop - there is no charge either way.
- 03day 1-2
Access and context
Repos, cloud accounts, Slack, issue tracker, and whatever architecture documentation exists. This is the step that decides whether week one is productive, so we ask for it early and chase it.
- 04day 2-3
First commit
The engineer picks up a small, real ticket - not a sandbox exercise - and opens a PR. It is deliberately something you can review quickly, because the fastest way to find a mismatch is to look at code.
- 05day 5-10
First feature
A scoped piece of the roadmap, shipped and reviewed by your team. By the end of the second week you have real code to judge the fit on.
IP and ownership
Everything the engineer produces is yours from the first commit - not on final payment, not on project completion. That covers source code, prompts and prompt templates, evaluation datasets and harnesses, fine-tuning data, infrastructure definitions, and documentation.
Work-for-hire
Work-for-hire assignment in the MSA, governed by US law.
NDA first
Mutual NDA signed before scoping, not after.
No residual licence
No residual licence for us, and no right to reuse your code in another engagement.
No training on your data
Your data is never used to train a model, ours or a vendor's.
What a US hire actually costs, next to this
The comparison that matters is not salary against monthly rate. It is fully-loaded annual cost against fully-loaded annual cost, including the months the seat sits empty. Here is the arithmetic for one mid-to-senior AI engineer.
| Line item | US hire | US contractor | MacroCoderz pod |
|---|---|---|---|
| Base salary or rate | $180,000 | $120/hr x 1,800 hrs = $216,000 | $2,800/mo x 12 = $33,600 |
| Payroll tax and insurance | ~$16,000 | None | None |
| Benefits, equipment, software | ~$22,000 | None | None |
| Recruiting fee (20% of base) | ~$36,000 in year one | None | None |
| Empty seat while hiring | 3-6 months of unshipped roadmap | 2-4 weeks | 48 hours |
| Ramp to productive | 4-8 weeks | 2-4 weeks | 2-5 days |
| Year-one total | ~$254,000 | ~$216,000 | $33,600 |
| Notice to stop | Severance and process | Per contract | 30 days |
Where the US hire genuinely wins: if this engineer will still be with you in three years, owns the architecture, hires and mentors the next two, and sits in the room for product decisions - hire. A pod is capacity and expertise, not a founding engineer.
$2,800 per engineer per month, 30 days notice to stop. Included: one full-time engineer, 40 hours a week; onboarding, handover documentation and an offboarding walkthrough; code review and architecture input from our senior engineers; daily written reporting; replacement at no charge.
Not included: model API spend, cloud infrastructure and third-party licences - these go on your accounts. Design, product management and QA as separate disciplines. Full rate card on the pricing page.
You do not need an engineering team to start
Describe the job you want done in plain words - answering customer questions, sorting incoming requests, pulling figures out of documents. On the first call we tell you whether AI is the right tool for it, roughly what it would cost and how long it would take.
If it is not worth building, we say so. If you are building a whole product rather than one tool, read how an AI product build works.
- Single-purpose agent$4,500-8,000
An assistant that answers your customers' questions from your own help pages and product documents, with a link to where each answer came from.
Typically 2-4 weeks - Production agent$8,000-18,000
An assistant that reads each incoming request, looks the customer up in your systems, drafts the reply and waits for a person on your team to approve it before anything is sent.
Typically 4-7 weeks
Frequently asked questions
Within 48 hours of the contract being signed, provided the role is one we have bench capacity for. If the brief needs a stack we have to source for, expect 5 to 10 working days. We tell you which case you are in before you sign, not after.
Yes. You get the engineer's CV and a technical call before anyone starts. If you want to run your own take-home or pair-programming exercise, we will do it - we would rather lose an hour than place someone your team does not want.
Tell us and we will propose a replacement, or give 30 days notice and we will wind the engagement down. We do not charge a replacement fee and we do not use a minimum term to make a swap difficult.
No. A dedicated engineer is full-time on your team and is not shared, time-sliced or rotated between accounts. That is the whole difference between this and a development agency's ticket queue.
You do, from the first commit. Code, prompts, fine-tuning data, evaluation suites, infrastructure definitions and documentation are all yours under the MSA. We keep no licence, no residual rights and no copy after offboarding.
Every engagement includes a written handover: architecture notes, runbooks, prompt and eval documentation, and a recorded walkthrough with your team. We do this in the last two weeks of the contract, not after it, so your engineers can ask questions while ours are still on the clock.
One month, rolling, with 30 days notice to stop. There is no annual contract and no minimum number of engineers. Most teams start with one engineer and add a second after the first sprint.
Start with one engineer
Tell us the stack and the roadmap item. We come back with a named engineer, a CV and a call.