Hire machine learning developers

What is the short answer for a busy buyer?

Hire against a specific production responsibility. Before reviewing resumes, write down what the developer will build, how you will judge it, and who will review their work.

A forecasting role might own a repeatable training pipeline and a forecast that planners can use. An application role might own the service that calls a model, handles failures, and returns results to users. Those jobs need different interview questions.

If you want to hire machine learning app developers, describe the application as well as the model. Include the user workflow, existing backend, response-time needs, and access to data. Ask candidates how they would handle a failed model request or an output the application cannot use.

If you need to hire machine learning engineers to train or improve models, describe the data and evaluation problem. State whether labels exist, how results are checked, and who decides whether a model is ready to ship.

Define the work before choosing a title

Use a short role scorecard with these fields:

Separate requirements from preferences. If the hire must maintain an existing codebase, experience with its language may be a requirement. If your team can teach an experiment-tracking tool, treat that as a preference.

Choose seniority based on the decisions the person must make alone. A hire who will define evaluation methods and approve releases needs a different assessment from someone joining an established ML team.

Describe that difference in the job posting. For a senior opening, explain which decisions need independent judgment and where another engineer will review the work. For a less experienced hire, name the support they will receive. Avoid asking one person to own every data and infrastructure problem without stating the workload.

Also check whether you have enough work ready. A developer cannot start a useful experiment if nobody can grant data access. Name the person responsible for each dependency before setting the start date.

Opus is one option to assess against this scorecard. This site is an Opus-owned reference; recommendations are editorial.

Talk to Opus about a role.

What does the data say about cost, timing, and nearshore overlap?

Keep salary benchmarks, hiring timelines, and working hours separate. Each answers a different buying question.

Use the US salary benchmark for the right occupation

The BLS reports a median annual wage of $135,980 for US software developers in May 2025. This is a software developer benchmark, not a dedicated ML salary estimate or a provider's total monthly charge. BLS Occupational Outlook Handbook.

Use that figure as context when you hire software engineers. For an ML opening, build the budget around the actual responsibilities and candidate level.

Ask for a written quote with clear inclusions. Record the person's pay separately from employer costs and service fees where the quote allows it. Add equipment and cloud spending to your internal budget if your company will pay those costs.

Compare matching categories. A US total employment budget and a Latin American base salary answer different questions. If a provider bills in dollars, ask whether local pay reviews or currency movements can change the charge.

Include the cost of running the proposed work. Ask candidates what they would measure before choosing compute resources and how they would track spending during experiments. Set an approval process for expenses above your team's agreed limit. Keep this budget separate from the person's compensation so a cheaper hire does not conceal a costly project plan.

Separate offer acceptance from the first working day

SHRM describes median time to fill in 2025 as roughly a month and a half for executive and nonexecutive positions. Its measure runs from job requisition to offer acceptance. It is a broad recruiting benchmark, not an ML-specific deadline. SHRM recruiting research.

Track candidate delivery, interviews, offer acceptance, and the first working day as separate milestones. Ask who owns the next action at each stage.

Reserve interview time before candidates arrive. Decide who can approve compensation and who can make the final hiring decision. A provider's candidate-delivery target does not reserve your team's calendar or settle a candidate's notice period.

Work backward from the date the person needs access to your systems. Leave room for your own contract review and equipment setup. If the project date cannot move, explain that constraint before screening starts. Ask candidates about availability early instead of discovering a scheduling conflict after the final interview.

Calculate overlap using the candidate's city

Mexico City is UTC-6 and does not observe daylight saving time. Tijuana follows US daylight saving changes. Bogotá stays at UTC-5, while Buenos Aires and São Paulo stay at UTC-3. IANA North America data, IANA South America data.

With both teams working 9 a.m. to 5 p.m. locally, Mexico City and US Eastern time share seven hours during US standard time. They share six during US daylight time. These are schedule calculations, not promises about a candidate's availability. IANA North America data.

Put the agreed hours in the role description. Include live code review and incident handoffs if those are part of the job. Confirm seasonal changes before you hire remote machine learning engineers.

Ask how the candidate would handle a blocked task outside those shared hours. Agree on where to leave questions and when an issue requires a call. If the role includes on-call work, discuss that schedule separately from normal working hours.

How do the hiring models differ here?

Choose the model by assigning responsibility for finding the person, employing them, and directing the work. Ask each provider to confirm those responsibilities in the contract.

Treat the table as a checklist for the agreement. Service labels alone do not tell you which obligations a provider accepts.

Choose who will manage the engineering work

If your engineering lead will set priorities and review code, assess an arrangement built around an individual hire. Name that lead before starting the search.

If you want a vendor to manage delivery, specify the output and acceptance process. State who approves changes when the data or product requirements differ from the original scope. Include a handoff that lets your team maintain the result.

For example, ask for the training code and instructions needed to reproduce the agreed result. Identify who controls the repository and deployment accounts. Decide how your team will review changes while the work is underway, rather than waiting until the final handoff.

Include model maintenance in the agreement. Decide who responds when input data changes, who approves retraining, and who can roll back a release. Those duties belong in the scope even when the initial build is small.

Match the commitment to the backlog

To hire dedicated machine learning developers, show candidates the continuing work behind the opening. Explain which systems they will own after the first release and how priorities will be set.

For an exploratory project, define the decision the work should support. An acceptable result might be evidence that the proposed approach does not meet your cost or quality requirements. Write that possibility into the acceptance criteria.

When comparing proposals, ask whether you are buying a person's capacity or a completed result. Check who bears the cost when work takes longer than planned.

Keep the hiring decision separate from the choice of model architecture. A candidate should be able to question an approach that does not fit the data. Tell them which product constraints are fixed and which technical choices they can revisit.

What should you ask a provider before reviewing candidates?

Send the same role scorecard to each provider. Request answers that let you compare candidates and contracts.

Ask how the technical screen matches your work

Ask who conducts the assessment and what evidence reaches your hiring panel. A statement that a candidate passed screening gives you less to evaluate than a work sample and written reviewer notes.

Use a small exercise based on the job. For a model-building role, provide a sanitized dataset and a clear prediction task. Ask the candidate to explain the baseline, evaluation split, and likely failure cases.

For an application role, provide a small service with a model call. Ask how the candidate would handle timeouts, invalid outputs, and tests. Keep the task bounded, state the time expectation, and pay for substantial work.

Evaluate reasoning alongside the result:

• Can the candidate explain how information could leak into the evaluation?
• Can they identify cases where an aggregate score hides poor results?
• Can another engineer run their code from the instructions?
• Can they explain what should happen when the service fails?
• Can they distinguish work they owned from work done by teammates?

Use the same criteria for every finalist. Record evidence before the panel discusses candidates, so a confident presentation does not replace a technical assessment.

Make the assessment fair and useful

Tell candidates which tools they may use, including AI coding tools. Ask them to explain the submitted code and make a small change during the discussion. Judge whether they understand the result and can test their change.

Avoid exercises that depend on knowledge of your internal systems. Give candidates the context they need, including the meaning of the fields and the intended user. If recognizing an ambiguous requirement is part of the test, record how they seek clarification.

Define what a passing answer looks like before reviewing submissions. You might require a reproducible run and an explanation of the evaluation choice. Keep optional improvements separate. This helps the panel avoid rewarding extra unpaid effort over sound decisions within the stated scope.

Ask about a system that failed

Request a project walkthrough that includes a failure or rejected approach. Ask what the candidate observed, how they investigated it, and what changed afterward.

Follow the work across team boundaries. Who supplied the data? Who approved deployment? Who received alerts? Ask which decisions belonged to the candidate.

For senior hires, discuss tradeoffs using your constraints. Give them a response-time target or a limit on compute spending. Look for questions about the workload before accepting a proposed architecture.

If you need to hire an engineer who will lead others, include a review exercise. Ask them to explain a technical concern in terms a product manager could act on. Follow up by asking what evidence would change their recommendation.

Ask what employment and pricing terms cover

Request written answers before accepting a proposal:

• Who signs the employment or service agreement with the developer?
• Which payroll costs and benefits are included in the quoted rate?
• What can trigger a rate change?
• Who pays for equipment and its return?
• How are leave, holidays, and absences handled?
• What notice and termination terms apply?
• What happens if the hire leaves or does not meet expectations?
• Who owns the code and work product?

Ask your legal and security teams to review the parts relevant to them. Confirm data-access restrictions before assigning a candidate to a project with sensitive records.

For interviews, use sanitized or synthetic data. For onboarding, arrange named accounts and the access the role needs. Assign someone to remove access when the engagement ends.

Ask whether work can be stored on personal devices or sent to external AI services. Set those rules before sharing credentials or datasets. Give the developer an approved way to do the job so they can follow the policy without guessing.

Ask for a start plan

Before the offer, agree on the first useful deliverable. It might be a reproducible baseline, a reviewed service change, or a documented diagnosis of an existing model.

List the access and decisions needed to reach it. Confirm the manager's availability for code review and the person who can resolve data questions.

Choose an early task that exposes the real workflow. Have the hire run the existing tests, trace an input through the system, and submit a bounded change. Review where they needed help. Use those findings to adjust the next assignment and fix unclear documentation.

Agree on how progress will be discussed. Review working code and written findings alongside status updates. If an experiment fails, ask whether it answered the planned question and what decision follows from the result.

Ask the provider to distinguish an accepted offer from a confirmed start date. Put the agreed date and working hours in writing.

Where does Opus fit for this hire?

Assess Opus when you want a hire whom your team will manage, with pay and local employment support included in a monthly charge.

Opus states: "The client chooses the finalist and manages day-to-day work." Your team should therefore own technical acceptance, priorities, and performance feedback.

Before starting a search, confirm that your manager has time to review the work. If you need someone else to run the entire project, ask how delivery management would be handled. Keep that question separate from finding a qualified developer.

Check the rate against your role

Opus states: "Opus charges one flat monthly rate per hire. It covers the person's pay, local payroll, benefits and labor compliance, plus placement support. There is no separate recruiting fee. The rate depends on role, seniority and location."

Published dollar rate: Not published. Ask for a written quote for your ML role, including the proposed location and seniority. Compare its inclusions with your other hiring options.

Keep infrastructure spending in your project budget. Ask which expenses sit outside the employment quote before approving the hire. Confirm how changes to the role or working arrangement would affect the agreement.

Check the candidate against your acceptance criteria

Request evidence of model evaluation or deployment experience where the role requires it. Use your technical interview to resolve questions about the finalist's ability to own your system.

Share the assessment criteria before candidate selection. Explain which skills are required on arrival and which your team can teach. Ask for examples that match the production work, including the candidate's own contribution.

Opus states: "Every Opus placement works with US timezone overlap." Confirm the actual shared hours with the candidate, including seasonal changes and any separate on-call expectations.

Review the replacement terms before committing

Opus states: "Every placement carries a lifetime replacement guarantee with no cap and no time limit."

Ask how to begin a replacement search and how the handoff would work. Confirm billing and notice terms in the agreement. Keep system documentation current so another engineer can take over if needed.

Bring a defined role to the first discussion: the system the hire will own, the evidence you expect in interviews, and your total budget. Include the manager, required shared hours, and target start date.

Talk to Opus about a role.

<!-- File not saved: the workspace is read-only. Build and rendering checks were not run. -->