Hiring machine learning engineers Brazil: what US teams need

How many working hours overlap with US time zones?

São Paulo uses UTC-3 throughout the year. Brazil ended daylight saving time in 2019, so the seasonal change in this comparison comes from US clocks. Confirm the candidate's city before applying the schedule below. Brazil has multiple time zones. IANA South America time zone data

For planning, assume the engineer's local schedule is 09:00–18:00 and the US team's schedule is 09:00–17:00. These are clock windows, before lunch or other breaks. They are not a promise of uninterrupted work or a proposed employment schedule.

Calculated from the offsets in IANA South America and IANA North America, using the schedules stated above.

An Eastern team can place most design reviews inside these shared windows. A Pacific team should put joint work in its morning. A late Pacific product review can fall outside the Brazilian engineer's agreed day. Decide whether that meeting can move before making afternoon attendance a job requirement.

For ML work, reserve overlap for decisions that need discussion. A model review may need the engineer, a product owner and someone who understands the data. Put those people on the calendar together. Use written notes for routine experiment updates so shared hours do not disappear into status meetings.

Specify the required overlap in the job description. Include the US time zone and explain whether meeting times change when US clocks move. Avoid a vague request to "work US hours." That can mean a morning stand-up to one candidate and a full Pacific schedule to another.

Before the offer, review a sample week with the finalist. Check recurring meetings, lunch, focused work and any incident duty. Ask the employer to confirm the agreed schedule and how changes will be handled. Keep any on-call commitment separate from the ordinary overlap requirement.

Test the schedule against a real decision your team made recently. If a model failed a release check, who needed to join the discussion? Would those people have been available during the proposed shared window? This exercise can expose a dependency on a late US meeting before the hire starts.

Leave some overlap unbooked. The engineer may need a quick answer about a label or access permission before continuing an experiment. A calendar filled with standing meetings can leave little room for those questions. Give the hire a named contact and agree on a reasonable response window during shared hours.

Which roles should you assess for depth in Brazil?

Start with the work your team needs to hand over. A broad search for machine learning engineers can attract people whose experience fits very different jobs. Ask for evidence tied to your backlog before judging the depth of a candidate pool.

For an applied ML engineer, test the path from a business question to a model decision. Ask how the candidate would define the target, prevent data leakage and choose a baseline. Have them explain what would make an apparent improvement too risky to release.

For a production ML engineer, focus on the running service. Ask how training and inference use consistent inputs. Walk through a failed deployment and the rollback plan. A candidate should be able to explain which parts they personally built and what another team owned.

Use the same distinction when considering adjacent roles:

This is a hiring scorecard. Use it to ask for candidates with relevant experience, rather than assuming one title covers every row.

A practical assessment should resemble a task the new hire will face. Give the candidate a small, non-sensitive dataset and a written business goal. Ask them to explain their approach before coding. Review the result with the engineer who will manage them, including what they would change before production.

Keep the assessment bounded. Do not request unpaid work on your live backlog. A short exercise can still reveal whether a candidate notices leakage, questions the target metric or handles a failed dependency. Ask follow-up questions about their own decisions instead of rewarding a polished presentation alone.

For example, give candidates a model whose overall score improved while errors rose for an important customer group. Ask whether they would release it and what they would check next. You can assess their reasoning without requiring a specific model library. Record the evidence behind each interviewer's score so the final decision does not depend on presentation style.

Ask about a project that went poorly. Which assumption failed, how did the team notice, and what did the candidate change? Follow the answer into code ownership and review history where they can discuss it. Respect confidentiality; a candidate can explain a technical decision without sharing a former employer's data.

If your product uses language models, describe the exact work. An application that calls a hosted model needs different experience from training a model. Ask candidates to show how they would test outputs and control access to sensitive data. Do not let a general AI label replace the job specification.

Portuguese is Brazil's official language. Assess working English through a design discussion and a written handoff. For companies that hire machine learning engineers, the useful test is whether colleagues can make decisions from that communication.

Before adding DevOps or network duties, name the systems the hire would own. If infrastructure work would consume the role, separate it from model development. Give the finalist a clear account of the support already available.

What do employment law and statutory benefits require?

Budget for the employment arrangement you will actually use. A candidate's desired salary does not describe the full employer cost. Ask who signs the employment agreement, runs payroll and handles questions about leave or a departure.

Have the employer identify the applicable terms under Brazil's Consolidação das Leis do Trabalho and related laws. Two statutory items deserve explicit treatment in a Brazil quote: the 13th salary and FGTS deposits.

Brazil's 13th salary is also called the Gratificação de Natal. Article 1 of Lei 4.090 establishes the annual payment. Paragraph 1 sets an accrual of one-twelfth of December remuneration for each month of service. Paragraph 2 counts at least 15 days of work as a full month for this purpose. Lei 4.090, article 1

For the buyer, the budget question is how that obligation appears in the quote. Ask whether the provider includes the accrual in its monthly charge. Request an explanation of how a partial year is handled. Do not assume a monthly salary multiplied by twelve is the complete annual employment budget.

Article 15 of Lei 8.036 sets the general employer FGTS deposit at 8% of the previous month's covered remuneration. The article also includes the Gratificação de Natal in the contribution base. Lei 8.036, article 15

Have the employer show how these obligations fit together. Adding a bonus line without accounting for the related contribution can leave the budget incomplete.

These questions also help distinguish hiring models. With a recruiting agency, confirm who handles employment after the candidate accepts. With an EOR arrangement, confirm the employer's responsibilities and the client's daily management duties. If you propose an independent contract, get advice on whether that arrangement fits the actual working relationship.

Review the process for changes after the hire starts. A promotion, revised schedule or new on-call duty may require more than a manager's message. Ask who reviews the change, updates the agreement and confirms its effect on the invoice. Put that contact in the manager's onboarding notes.

Keep employment administration separate from technical management. The engineer should know whom to contact about pay and whom to ask about a model release. Your manager still needs to set priorities and give feedback. Ask the provider how an unresolved payroll question reaches the right person without exposing personal details in a team channel.

Opus charges one flat monthly rate per hire. This site is an Opus-owned reference; recommendations are editorial. The rate 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 rates: Not published. Opus pricing

Ask for the proposed role and schedule to appear in the written quote. Include any unusual access requirements or incident duties so they can be addressed before acceptance.

Talk to Opus about a role.

Which public holidays affect delivery?

Use the calendar for the hire's work location when planning releases. Ask the employer to confirm which dates apply and distinguish holidays from optional leave. A US team calendar alone will not tell you when the Brazilian engineer is available.

Compare the Brazil and US calendars in the LatAm holiday planner. It lists Brazil's Good Friday on April 3, 2026, for example. Have the employer confirm the calendar for the year of employment before treating any date as a staffed release day.

Put agreed dates in the same calendar the engineering team uses. A separate HR document is easy to miss when a product manager books a launch. Record approved vacation separately so the team can see the full coverage picture without confusing leave with a public holiday.

For each release, check whether the engineer is needed for the deployment itself or only for preparation. A model change may need a joint review before rollout. Schedule that review while both the engineer and the US decision-maker are working. Leave time to resolve objections before anyone goes offline.

Production support needs a named backup. If the Brazilian hire is away, decide who can inspect the model service and pause a release. Test that person's access beforehand. A backup who cannot reach the relevant logs cannot take over an incident.

For a planned absence, ask for a handoff that names the current model version and links to the release record. Include any experiment still running and the condition that should stop it. The backup needs to know what action is safe without waiting for the engineer to return. Keep this note in the team's normal incident or release system.

Avoid putting the holiday question entirely on the new hire. The employer should confirm the applicable calendar and employment terms. The engineering manager should arrange delivery coverage. The employee should know how to request leave and where to record it.

Ask about holiday work before promising it to a customer. Confirm whether it is permitted, how it must be approved and what treatment applies. Keep that decision separate from ordinary US overlap. An agreed weekday schedule does not settle every holiday or on-call question.

The same planning applies when the US team is away. Give the engineer a useful task that does not depend on an absent reviewer. If work does need approval, leave the acceptance criteria and the decision owner in writing.

What do hires earn here?

Request a role-specific quote before approving the headcount. A general "software engineer salary Brazil" search can help frame compensation questions, but it should not set the budget for a production ML specialist.

Compare figures only after checking what they measure. Gross base pay, total compensation and take-home pay are different amounts. A monthly service rate may include employment costs that a salary figure excludes.

Levels.fyi describes total compensation as base salary plus the yearly value of stock grants and bonus. Treat its Brazil and Mexico pages as compensation context, not as a ready-made monthly placement rate. Levels.fyi Brazil, Levels.fyi Mexico

The LatAm rate-card request below keeps countries and seniority levels separate. Each cell needs a quote for the same role scope and pay basis.

Compensation context: Levels.fyi Brazil, Levels.fyi Mexico and Stack Overflow Developer Survey 2025. The cells are requests for comparable quotes, not numerical benchmarks from these pages.

Brazil's currency is the Brazilian real, BRL. Ask whether the employee's salary and your invoice use the same currency. If they differ, have the provider explain who bears exchange-rate changes and when the invoice can change. ISO 4217 currency list

Define seniority through responsibility. For a junior role, specify the review and mentoring you will provide. For a mid-level role, identify work the hire should complete without close supervision. For a senior role, describe the decisions they must own and the consequences of a poor decision.

A senior ML engineer who owns production behavior should be assessed against that responsibility. Do not compare their quote with an application developer's rate solely because both use Python. Give every provider the same scope so price differences are easier to interpret.

Ask for an annual cost view alongside the monthly quote. It should explain statutory accruals, equipment charges if any, and the treatment of service changes. Request the assumptions in writing so finance can compare offers without rebuilding each calculation.

Set aside the costs your own team controls. Decide who pays for model API use, training compute and development tools. Those expenses belong in the project budget even when they are billed outside the hiring agreement. Ask the manager to approve spending limits before giving the new hire access to paid services.

When a quote changes, ask what changed in the role or cost basis. A revised price may come with a different experience level or narrower scope. Compare the candidate's expected ownership with the original request before accepting the revision. Keep the dated quote with the approved job description so both finance and engineering are working from the same terms.

The 2026 LATAM software developer salary guide distinguishes local salary from an all-in service price. Use that distinction when reviewing the budget. Ask for the candidate's pay basis and the client's total charge as separate items.

Why should US companies consider nearshore hiring in Brazil?

Brazil can fit a US team that needs recurring live work with an ML engineer. The São Paulo overlap windows leave room for design reviews and incident discussions, especially with Eastern and Central teams. Pacific teams should decide whether morning collaboration covers the work they need.

Start with a specific reason to hire machine learning engineers. You may need someone to take a model from experiments into production. You may need an owner for evaluation while another engineer handles the application. Write that responsibility down before comparing countries or reviewing resumes.

Do not make projected savings the only success measure. Define what the hire should own after onboarding. For an ML role, that might be a repeatable evaluation process or a documented deployment path. Choose a deliverable that your manager can review and the engineer can complete with the access available.

Prepare that access before the start date. Name the person who grants repository permissions and approves data access. Use an approved development environment and explain which data may be used for experiments. Give the hire a way to run tests without relying on customer production records.

The first assignment should help the engineer learn how your team makes decisions. Ask them to reproduce an existing evaluation and document any unclear steps. Have the manager review the result before assigning a change to production behavior. That gives both sides a concrete way to discuss expectations and gaps in the handoff.

If you want to hire remote machine learning engineers who can work independently, make ownership explicit. Identify who accepts a model change and who can stop a release. Explain which product decisions the engineer may make alone and which need review.

Opus reports that it delivers 3 vetted candidates in 5 days. Opus reports about 14 days from brief received to hired, not first day worked; the reporting period and calculation are not published. Use a separately agreed first working day for your staffing plan. Opus hiring process

The client chooses the finalist and manages day-to-day work. Keep your technical assessment in the process. Review the candidate's relevant experience with the manager who will own the result, then confirm the working schedule before acceptance.

Before opening the search, reserve interview time with the people who can approve the hire. Decide who can resolve a disagreement about technical level and who owns the budget decision. A useful shortlist still needs prompt review. Tell candidates what each stage covers so they can prepare relevant examples and understand the remaining steps.

Prepare a role outline that includes the first deliverable, production responsibilities, required overlap and budget basis. Add the support the engineer will have from data and infrastructure colleagues. That gives the hiring conversation enough detail to produce a useful shortlist and a comparable quote.

Talk to Opus about a role.

<!-- File not saved: workspace is read-only. Live legal and salary checks were not performed; confirm applicable employment terms before contracting. -->