Fractional Lead AI Engineer

AI Teams Builder

by Ahmed Mostafa

You hired AI engineers. They're shipping demos that work on stage and break in production. I step in for three to six months as the Lead AI Engineer your team is missing — and leave once they can ship without me.

Services

Four ways I help your team ship.

Process

From the intro call to the handover.

Six steps. Every engagement runs the same shape — only the dates, the scope, and the success criteria change.

  1. The intro call

    30 minutes

    You describe the situation — your team, what's working, what isn't, what you're trying to ship. I tell you whether I'm the right fit. No slides, no obligation. If it isn't a fit, I'll say so on the call.

  2. Scoping & agreement

    Within one week

    If we're a fit, we write the engagement together. Scope, milestones, success criteria, weekly cadence, and the exit date — on paper before either of us signs. You get a fixed monthly fee and a fixed end date.

  3. Week one — the assessment

    Days 1 – 5

    I join your team. I read your code, sit in your standups, and talk to every engineer 1:1. By Friday, you have a written technical assessment: what's working, what's at risk, the three things we do first.

  4. Lead & enable

    Months 1 through 5

    I take the Lead AI Engineer seat. Architecture calls, hiring loops, code review, weekly 1:1s, MLOps standards, and the production discipline that turns demos into systems. I'm in your Slack, your sprint reviews, your roadmap calls.

  5. Handover

    Final two weeks

    One of your engineers steps into the Lead AI Engineer seat while I shadow them. Everything still in my head — decisions, rationale, the playbook — moves into a written runbook your team owns after I'm gone.

  6. Exit, and after

    Day after, indefinitely

    I'm gone. Your team is making the calls without me — that was always the point. If something breaks six months later, the first call back is on me. No retainer, no lock-in.

Methodology

Four phases, with the exit built in.

  1. Assess

    Week one, I listen — to your engineers, to your code, to your business goals. By Friday, you have a written assessment: what's working, what's at risk, what we do first.

  2. Lead

    Then I take the Lead AI Engineer seat. Architecture choices get made. Standards get written. The team has a direction — and someone accountable for it.

  3. Enable

    While I lead, I'm teaching. Code reviews, 1:1s, and the production discipline that turns demos into systems your customers can rely on.

  4. Exit

    When your team is making the calls without me, I leave. Not a surprise, not a renegotiation — the exit was on the calendar from day one.

Stack

I help teams pick the right tools.

The AI tooling ecosystem changes every month. Most teams chase the new thing. I help yours choose the tools that fit your actual problem, ship with them in production, and walk away from the ones that don't earn their keep.

LLMs & agents
LangChain, LangGraph, OpenAI, Anthropic, HuggingFace
LLMOps
LangSmith, LangFuse, LangWatch
ML / training
PyTorch, Transformers, Fine-tuning toolchains
Infra & MLOps
Kubernetes, Docker, Airflow, GitHub Actions
Cloud
AWS, GCP, Azure, SageMaker, Vertex AI

About

Fourteen years in software. Eight in AI.

I spent fourteen years writing software — the last eight on AI. Most recently I was CTO of an enterprise AI startup, which is to say I've sat on both sides of this engagement. I've been the founder who needed AI leadership. I've been the Lead AI Engineer who came in. MSc in Data Science, Sapienza University of Rome.

Today I work fractionally — short, deep engagements where the point isn't to stay forever, it's to leave your team better than I found it. I also run llmsengineering.com, where I take on hands-on RAG and agent builds for teams that need delivery rather than leadership.

Based in Rome. Three to six month engagements. Remote, with teams worldwide.

Contact

Ready to talk?

Tell me about your team — what's working, what isn't, what you're trying to ship. I read every inquiry myself and reply within 48 hours. Next steps if we're a fit, an honest no if we aren't.