Tech hiring Data & AI
Hire AI engineers who get models into production.
An AI engineer builds product features on top of large language models: assistants that answer from your own documents, agents that carry out multi-step tasks, pipelines that read and classify what people send you. The skill is less in calling a model than in making its output reliable, measurable and affordable.
01 The role
After the demo, the engineering starts.
An AI engineer takes a foundation model that someone else trained and turns it into a feature your customers or staff can rely on. That means choosing the model, designing the prompts and the context it receives, connecting it to your data and tools, and wrapping the whole thing in ordinary, well-tested software.
The hard part comes after the demo. A prototype that answers ten questions well is easy; a system that answers ten thousand, says “I do not know” when it should, stays inside a cost and latency budget, and does not leak data is engineering. Good AI engineers build an evaluation set early and treat every prompt or model change as something to be measured, not eyeballed.
This is a different job from machine learning engineering. An ML engineer trains, tunes and serves models from your own data. An AI engineer mostly composes existing models with retrieval, tools and guardrails. Many products need the second long before they need the first.
What they ship
- Assistants that answer from your documents, tickets or knowledge base (RAG)
- Agents that call tools and APIs to complete multi-step tasks
- Extraction and classification pipelines for emails, forms, contracts and calls
- Semantic search across products, content or internal records
- Evaluation suites that score answers before a change goes live
- Guardrails: input filtering, output validation, permission-aware retrieval
- Cost, latency and quality monitoring for model calls in production
02 Skills and stack
What AI engineers work with, and what is optional.
The tooling column changes faster here than in any other role we hire for. Weigh the core column most heavily: it is what transfers when the libraries change.
Core
The role cannot be done without these.
- Python, written as production software
- LLM APIs: prompting, structured output, tool calling
- Retrieval-augmented generation: chunking, embeddings, re-ranking
- Evaluation: test sets, graders, regression checks
- API and backend engineering
- Data handling and privacy
- Reasoning about cost, latency and failure modes
Tooling
Varies from team to team.
- Model providers: Anthropic, OpenAI, Google, open-weight models
- Vector search: pgvector, Pinecone, Weaviate, Qdrant
- Orchestration: LangGraph, LlamaIndex, or plain code
- Model Context Protocol (MCP) for tool integration
- FastAPI, queues and background workers
- Tracing and evaluation: Langfuse, LangSmith, Braintrust
- Hosting: AWS Bedrock, Azure AI Foundry, Vertex AI
- Docker and CI/CD
Adjacent
Useful, not required.
- Fine-tuning and applied machine learning
- Natural language processing fundamentals
- Computer vision and speech models
- Data engineering and pipelines
- TypeScript for chat and streaming interfaces
- Statistics, for reading evaluation results
- Ethics and bias awareness
03 When to hire one
When you need an AI engineer, and when a developer with an API will do.
Plenty of AI features do not need a dedicated AI hire, and some problems are not language-model problems at all. Check both lists before you write the job description.
Hire an AI engineer when
-
A prototype impressed everyone and then stalled
Someone built a demo in a week. Months later it is still not live because nobody can say how often it is wrong. That gap is exactly what an AI engineer closes.
-
People keep asking the same questions of the same documents
Support, sales or operations staff spend hours searching policies, manuals or past tickets. Retrieval over your own content is the most dependable first use of a language model.
-
A manual review step is limiting how much you can process
Reading and routing emails, checking forms, summarising calls: structured extraction with a human checking the exceptions can remove the bottleneck.
-
You are adding AI features to an existing product
The feature has to respect your permissions, your data model and your uptime. You need an engineer, not a prompt.
-
Model costs are rising faster than usage
Caching, routing simple requests to smaller models and trimming context are engineering work, and often pay for the hire.
Look elsewhere when
-
You need a model trained on your own data
Forecasting, scoring, recommendations and custom vision models are machine learning engineering: training, feature pipelines and model serving.
Hire machine learning engineers -
Your data is scattered, undocumented or wrong
An assistant built on bad data gives confident bad answers. Get the pipelines and the warehouse in order first.
Hire data engineers -
The question is “what does our data say?”
Finding patterns, running experiments and explaining results is analysis, not product engineering.
Hire data scientists -
You want a chat widget on an otherwise conventional app
A capable backend developer can integrate a hosted model API. Hire an AI engineer when quality, evaluation or scale become the problem.
Hire Python developers -
Nobody has decided what the AI feature is for
Start with a short, defined pilot before committing to a permanent hire.
Post a freelance project
04 How we assess
How we assess AI engineers, before you meet one.
A CV says what someone claims. We test what they can do, then send you the evidence with the shortlist.
-
Stage 1
Profile review
We look for AI systems that reached real users, not notebooks and course certificates: what the system did, which models and retrieval approach it used, how it was evaluated, and what the candidate personally built.
-
Stage 2
Technical test
A timed test covering Python and software fundamentals, how language models and embeddings behave, and a practical task such as designing a retrieval pipeline or getting reliable structured output from a model.
-
Stage 3
Conversation with our team
Our engineers ask the candidate to walk through a system they shipped: how they knew it was good enough to release, what it cost per request, how it failed and what they did about it. We listen for measurement, and for honesty about limits.
-
Stage 4
Review and levelling
Reviewers score accuracy, speed, problem-solving, communication and teamwork, then assign a level using our AI framework. The notes separate people who have run AI features in production from those who have only built prototypes.
-
Then
Your shortlist
Assessed candidates, with notes and scores.
The engine
Evalia
Assessments run on Evalia, the assessment platform we built ourselves. Across the ProDevs network, more than 100,000 assessments have been completed.
The levels we assign are public. Read the AI framework, or see how a hire runs from brief to offer.
05 Seniority
How much can you hand over? It depends on the level.
The title on a CV tells you little. What matters is how much an AI engineer can be handed without someone checking each step.
Junior
Owns a well-scoped component
- Implements a prompt, a retrieval step or an evaluation script to a clear specification
- Usually a software developer or data graduate moving into AI work
- Needs a senior colleague to judge quality and catch subtle failure modes
- Rarely the right first AI hire for a team with no one to learn from
Mid-level
Owns a feature in production
- Builds a retrieval or agent feature end to end, with tests and an evaluation set
- Chooses a model and a retrieval strategy and can show the numbers behind the choice
- Monitors quality, cost and latency after launch and acts on what they see
- Explains to non-specialists what the system can and cannot be trusted to do
Senior
Owns the AI platform and its risks
- Designs the shared pieces: retrieval, evaluation, tracing, model routing, guardrails
- Decides build, buy or fine-tune, and when not to use a language model at all
- Sets the release bar and the review process for anything customer-facing
- Works with legal and security on data handling, and advises leadership on what is realistic
Indicative salary
Data and ML engineering band. Annual, in US dollars, for remote roles with US, UK and EU companies.
- Mid-level
- $35–65k
- Senior
- $65–100k
- Typical US equivalent
- $140–220k
Indicative market ranges as of August 2026, based on ProDevs placement experience and published market data. The exact figure depends on stack, seniority and time-zone requirements. See the full cost breakdown.
06 Interview questions
Interviewing an AI engineer? Ask these.
The field moves quickly, so these questions test judgement and measurement, not knowledge of this year’s libraries.
-
Tell me about an AI feature you shipped. How did you know it was good enough to release?
A good answer shows: A concrete evaluation: a test set drawn from real cases, a defined pass mark, and a named person who agreed it. “We tried a few prompts and it looked good” is the answer you are screening out.
-
A retrieval-based assistant gives a wrong answer. How do you work out why?
A good answer shows: A split between retrieval failure and generation failure, and a way to tell them apart: inspect what was retrieved, check chunking and ranking, then the prompt. Candidates who reach straight for a bigger model have not debugged one.
-
When would you fine-tune a model instead of improving prompts and retrieval?
A good answer shows: A reluctance to fine-tune first, with reasons: cost, maintenance, data requirements. They can name the cases where it does help, such as a narrow format, a fixed style or lower latency on a smaller model.
-
How do you stop an assistant from revealing data a user should not see?
A good answer shows: Permissions enforced at retrieval, not requested in the prompt. Awareness of prompt injection through documents and tool output, and of logging what the model was shown.
-
How would you halve the cost of a feature without users noticing?
A good answer shows: Practical levers: prompt caching, shorter context, routing easy requests to a smaller model, batching. And an evaluation run to confirm quality held, because the saving is worthless otherwise.
-
When is an agent the wrong design?
A good answer shows: Judgement that a fixed workflow is cheaper, faster and easier to test when the steps are known. Agents earn their place when the path cannot be predicted. Over-use of agents is a common sign of inexperience.
-
What changed in your approach over the last year, and what made you change it?
A good answer shows: Evidence that they follow the field critically: they dropped a framework, changed how they evaluate, or moved to a newer model for a stated reason, not because it was fashionable.
07 Ways to engage
Not only permanent hires. Four ways to work with an AI engineer.
A permanent hire is the usual route, not the only one. Choose by how long the work lasts and who should employ the person.
- Placement fee, of first-year salary. One-time.
- 10–15%
- Replacement guarantee on direct hires.
- 2 weeks
- To a first shortlist, for most roles.
- 48h
-
Direct hire
A permanent employee on your contract and your payroll. One-time placement fee of 10–15% of first-year salary.
Tech hiring -
Dedicated team
An AI engineer inside a team we assemble and run for you: product, design and engineering on your roadmap, for a monthly retainer.
Dedicated teams -
Outsourcing & payroll
A full-time person in a country where you have no legal entity. ProDevs employs and pays them: 20% of salary if we find the person for you, or 5% of salary, capped at $500 a month, for payroll and compliance on someone you have already chosen.
Outsourcing -
Freelance project
A defined piece of work with a clear end. You pay the agreed price per milestone through escrow, plus a 5% client fee. Free to post.
Freelancers
08 Around the role
Around the role: who an AI engineer works with.
An AI engineer is a software engineer first. They work inside a product team, and the quality of what they build depends heavily on people who know the subject matter and can say what a correct answer looks like.
- Product manager
- Agrees what the feature must get right, how often, and what happens when it is unsure.
- Subject-matter experts
- Supply and review the examples that become the evaluation set.
- Backend developers
- Integrate model calls with existing services, permissions and data.
- Data engineer
- Provides clean, current, access-controlled sources for retrieval.
- Security and legal
- Review what data reaches a model provider and what is retained.
09 Client testimonials
Hear it first-hand, from companies that hired.
Founders, operators and executives describe working with ProDevs, in their own words.
Play video: Client testimonial: Olufemi Seyi, COO, Casafina development company
Play video: Client testimonial: Aproko Doctor, Building AwaDoc
Play video: Client testimonial: JT Liddell, Promenade CEO
10 Questions
Before you hire an AI engineer: the usual questions.
Something else? Book a call (opens in a new tab) and ask us directly.
-
What is the difference between an AI engineer and a machine learning engineer?
An AI engineer builds product features on top of existing foundation models, using prompting, retrieval, tool calling and evaluation. A machine learning engineer trains, tunes and serves models from your own data. If you want an assistant that answers from your documents, you need the first. If you want a fraud score or a demand forecast, you need the second.
-
What does an AI engineer need to know in 2026?
Solid Python and backend engineering, how to get structured and reliable output from language models, retrieval-augmented generation, tool calling and agent design, and above all how to evaluate a system so that changes are measured. Familiarity with the Model Context Protocol and with tracing tools is increasingly common.
-
How do you assess AI engineers when the field changes so quickly?
We assess the parts that last. The four stages are a review of systems the candidate has shipped, a timed technical test, a conversation with our engineering team and a final review that assigns a level. Throughout we look for software fundamentals and a habit of measuring quality, not for a list of this year’s libraries.
-
How quickly can I hire an AI engineer?
For most roles you see a first shortlist within 48 hours of agreeing the brief, and most companies find their hire within two to three weeks. Senior AI engineers are in demand, so a clear brief and a short interview process make a real difference.
-
Do we need an AI engineer, or can our developers use an API?
For a simple feature, such as summarising a record or drafting a reply, a good backend developer with a hosted model API is often enough. Hire an AI engineer when answers must be grounded in your own data, when mistakes are costly, or when cost and latency at volume start to matter.
-
Can I hire an AI engineer for a pilot project first?
Yes. You can post a defined project and work with a vetted freelance AI engineer, paying per milestone through escrow. A pilot with a clear success measure is a sensible way to find out whether a permanent hire is justified.
Talk to ProDevs
Tell us about the role.
Tell us the level and the stack. We will tell you how quickly we can put assessed AI engineers in front of you.
Project work
Need it for a project instead?
Not every job needs a full-time hire. Describe the work, get proposals from vetted freelance AI engineers, and pay per milestone through escrow. Free to post.