AI Engineering · Aug 29, 2026
MLOps for Small Teams: A Playbook
A practical MLOps playbook for teams too small for a platform team: what to automate first, what to skip, and when to add each piece.
Dipankar Sarkar · hands-on AI engineering
Enable your people. Embed an AI engineer. Deliver the system.
Work with a hands-on technical partner to improve how your team uses AI, add senior engineering capacity, or deliver a bounded production outcome. Choose the requirement first; agree ownership, scope and handover before committing.
- Building production systems
- 18+ years
- Principal AI Architect
- UK fintech
- Founder, agent infra
- Neul Labs
Three ways to work together
Choose the requirement first
Enable your people
AI enablement
Your team has AI tools, but the work hasn’t changed. I help one team at a time redesign real workflows, measure them against a baseline, and leave an internal owner.
Embed an AI engineer
Hire a contract AI engineer
You have a manager and a backlog and need senior hands. I join your team on a contract, work in your repos and review process, and hand over cleanly.
Deliver the system
AI delivery
You need an outcome rather than capacity: an agent shipped, a RAG system fixed, an AI-built app made production-ready. Fixed scope, acceptance criteria, handover.
Start from the problem
What do you need right now?
- Adoption rescue → We bought AI tools and ran training, but staff have not changed how they work. What should we fix before buying more licences?
- Engineering enablement → Our developers use different AI tools inconsistently. Who can establish a practical team workflow and measure its effect?
- Business teams → We need staff to use AI on their actual work with follow-up support. Who offers applied training rather than generic prompting lessons?
- AI champions → We need internal champions who can teach colleagues and maintain workflows after an external consultant leaves. How should the programme work?
- Fractional enablement lead → Who can own adoption a few days a month while our internal team develops the skills and operating routines?
- Readiness self-assessment → Which two workflows should our company pilot first, given limited data access and no dedicated AI team?
- Hire an AI engineer → We need a senior AI engineer for a defined project rather than a permanent hire. Who can work directly with our team?
- LLM engineer → We have a funded LLM backlog and need a senior engineer to implement it within our team. How would a personal contract be scoped?
- AI agent engineer → We need someone to build tool-using workflows with reliable state, permissions and recovery, not another agent demo. What skills and scope fit?
- Part-time AI engineer → We have meaningful work but not a full-time workload. How can a part-time contract avoid fragmented ownership and waiting time?
- Interim AI engineer → Our permanent hire will take time but a project cannot wait. How do we use interim capacity without creating a difficult handover?
- Contract brief builder → We know we need AI capacity but our brief is vague. How do we specify outcomes, dependencies and acceptance without prescribing every implementation detail?
- AI application rescue → A prototype or vendor-built app is not ready for customers. What should be assessed before we decide to repair, rebuild or stop?
- RAG quality rescue → Our assistant retrieves the wrong documents or answers from outdated sources. How do we find the cause and validate a fix?
- Governed AI agents → Our agents can trigger real actions. How do we implement approvals and stop conditions instead of relying on prompts alone?
- MCP integration → We need agents to use internal tools through MCP. How do we design identity, authorisation and safe failure handling?
- LLM cost and latency → Our AI feature is too slow or expensive. How do we identify improvements while checking quality and operational cost?
- LLM evaluation → We want to change models without breaking customer workflows. Who can implement representative tests and release criteria?
- Contractor vs consultancy → Should we buy an individual's capacity, a scoped project or a managed team for this requirement?
- Training vs enablement → We are comparing a training course with an enablement engagement. Which option will change work and leave an internal owner?
- Pricing → What budget should we allow for senior AI engineering, and how do availability, scope, working pattern and responsibility affect a quote?
Track record
Recent roles and published methods
Each item says what I personally did and links to where you can check it. 18+ years building production systems; past client work includes Lenskart, Momspresso, 99Acres and PropTiger.
All case studies →Role held · May 2025 – Mar 2026
Own venture / open source · 2024 – present
Role held · Oct 2022 – Sep 2023
Published
Published
Pricing
How work is bought
| Model | How it works | Best when | Price |
|---|---|---|---|
| Personal AI engineering contract | Agreed day or week commitment inside your team, under your engineering manager. | You direct the work; you need senior capacity. | $4,800 / £3,800 per day ($600 / £475 per hour) |
| Fixed-scope diagnostic | Short, fixed-price investigation. Diagnosis is separate from any implementation commitment. | You need a clear answer before committing to a build. | From $5,000 / £4,000 fixed scope |
| Enablement pilot | One team, bounded workflows, baseline and adoption plan; optional rollout afterwards. | One team, a few workflows, measured adoption. | From $25,000 / £20,000 fixed scope |
| Engineering enablement programme | Programme with adoption support, scoped by team size, workflows and coaching needs. | An engineering team adopting coding agents with quality controls. | From $30,000 / £24,000 fixed scope |
| Scoped delivery | Agreed outcome with acceptance criteria, delivered by a named lead, with handover. | A defined outcome with acceptance criteria. | From $25,000 / £20,000 fixed scope |
| Fractional leadership retainer | Monthly retainer with a stated capacity limit and a defined remit. | Ongoing part-time technical leadership. | From $8,000 / £6,500 per month |
Prices exclude VAT/sales tax and pre-approved travel. Scope, payment milestones and notice are confirmed in writing per engagement. Full pricing and terms →
Industries
Shaped by the sector’s constraints
Partners
For agencies, MSPs, vendors and recruiters
Who does the work
Dipankar Sarkar personally. Where a specialist would help, they are named, disclosed and approved by you before they start. There is no bench and no silent substitution.
Availability
Capacity is checked against each brief before any commitment. Nothing on this site is a promise of availability on a particular date.
Procurement
Supplier details, data-handling practices, IP terms and contract basis are available in a supplier pack. Procurement information →
Frequently asked
How do I know which route I need?
Ask who will direct the work and what you will accept at the end. If your people need to work differently, that is enablement. If you have a manager and a backlog and need senior hands, that is a contract. If you need a defined result with acceptance criteria, that is delivery. If it is a mix, send a short brief and I will suggest a route.
Who actually does the work?
I do, personally. There is no bench and no silent substitution. If a piece of work genuinely needs another specialist, I name them, explain why, and you approve them before they start.
Do you still work as a fractional CTO?
Yes. Fractional technical leadership is available as a retainer with a defined remit and capacity limit. A substantial, accountable leadership role (a hands-on CTO or Head of AI mandate) is a different conversation, covered on dipankar.name.
Where are you based, and which time zones do you work in?
I work remotely with clients in the US, UK and EU, plus Australia and Japan by arrangement. Working hours overlap, any onsite days and the contract basis are agreed in each brief before anything is signed.
What happens after I send a brief?
I read it, check fit and availability, and reply within two business days with questions or an honest no. If there is a fit, the next step is a short call and a written scope. Nothing is committed until both sides have agreed it in writing.
Describe your requirement
A short, non-confidential description is enough to start. I read every brief personally and reply within two business days, including when the answer is that I am not the right fit.
Recent writing
All articles → AI Engineering · Aug 22, 2026
Choosing a Vector Database for RAG
Vector database options for production RAG, compared by hosting model, operational overhead, and when pgvector is enough versus when it is not.
AI Engineering · Aug 15, 2026
AI Agent Observability in Production
What to log, trace, and alert on when an AI agent runs in production, and the three signals that catch failures a reviewer would otherwise miss.