Job description template

Machine Learning Engineer

You will design, build, and maintain machine learning systems that drive measurable product and business outcomes. As a mid-level engineer, you will work independently on well-defined work, partnering with product and engineering to turn requirements into reliable models and production-ready pipelines.

Responsibilities

Develop and iterate ML models (e.g., ranking, classification, regression, or forecasting) based on product goals and data availability
Build and productionize ML data pipelines and training workflows with reproducibility and monitoring in mind
Integrate model outputs into downstream services or batch/streaming workflows used by product teams
Evaluate model performance using offline metrics and online/production signals, and document trade-offs clearly
Monitor model quality and drift, investigate issues, and implement fixes to maintain reliability
Collaborate with data engineering and software engineering to improve data quality, feature reliability, and system performance

Must-have qualifications

Hands-on experience training and evaluating ML models in production environments
Experience designing end-to-end ML pipelines with reproducibility (data, features, training, and model versioning)
Proficiency with Python and common ML/data libraries for model development and experimentation
Strong understanding of model evaluation, error analysis, and metric selection
Experience implementing monitoring for model quality and drift and responding to model degradation
Ability to integrate ML models into services or batch workflows using standard software engineering practices

Nice to have

Experience with MLOps tooling and deployment workflows (e.g., CI/CD for models, model registry, or experiment tracking)
Background in specific domains such as personalization, fraud/risk, search/recommendations, or forecasting
Experience working with big data or streaming/batch systems for large-scale training and inference

What success looks like in 12 months

Increase model quality on priority use cases
Ship production ML workflows end-to-end
Reduce model incidents and degradation impact
Establish effective monitoring and drift response
Improve experimentation velocity and reproducibility

Logistics

Full-time (40 hours a week), remote from Latin America.
Working hours overlap with US business hours.

About the company

[About your company: what you build, your stage and team size]

Want it tailored to your stack, seniority and time zone?

Customize with the free generator →

Interview scorecard

Score every candidate against the same bar

Deliver production-ready machine learning systems that improve product outcomes through reliable modeling, robust pipelines, and continuous monitoring.

Outcomes to measure

What this hire should deliver in the first year. Replace [N] with your own targets.

Increase model quality on priority use cases. Measure: improve relevant offline metrics by [X]% for [1-3] priority models
Ship production ML workflows end-to-end. Measure: deliver [N] end-to-end ML pipelines/models integrated into production or operational workflows
Reduce model incidents and degradation impact. Measure: reduce production model-related incidents by [X]% and achieve [Y]% successful model runs
Establish effective monitoring and drift response. Measure: implement monitoring that detects quality/drift issues with alert coverage of [Z]% and reduce time-to-diagnosis by [X]%
Improve experimentation velocity and reproducibility. Measure: reduce time from experiment to reproducible training run by [X]% and increase experiment reproducibility pass rate to [Y]%

Competencies and weights

What to evaluate, how much it counts and what a strong answer looks like.

Production model development · 22%
The role requires training, evaluating, and iterating models in environments where reliability matters. Evaluated in Round 1.

Weak: Describes model work only in offline terms and cannot explain how production constraints were handled.
Meets the bar: Explains model development with at least one production context, including evaluation approach and iteration steps.
Strong: Demonstrates clear end-to-end thinking from offline validation to production performance, including handling failure modes.

End-to-end ML pipelines · 18%
Independent work on well-defined tasks requires building reliable pipelines with reproducibility and versioning. Evaluated in Round 1.

Weak: Relies on manual steps and cannot describe how training data/features are reproduced.
Meets the bar: Describes an ML workflow with reproducible datasets/features, versioning, and restartability.
Strong: Shows robust pipeline design including data/feature integrity checks, repeatable training, and well-defined artifacts.

Evaluation and error analysis · 18%
Correct metric selection and diagnosing errors are key to improving model outcomes. Evaluated in Round 2.

Weak: Uses metrics without explaining why they match the objective; limited discussion of error analysis.
Meets the bar: Explains metric choice, baseline comparisons, and includes structured error analysis methods.
Strong: Connects evaluation to business/product impact and describes targeted diagnostic work that changes model decisions.

Monitoring and drift response · 18%
The role must maintain quality over time and respond effectively to degradation. Evaluated in Round 2.

Weak: Mentions monitoring in general terms and cannot explain thresholds, signals, or response workflow.
Meets the bar: Describes monitoring signals for quality/drift and how alerts connect to investigation and remediation.
Strong: Shows a mature monitoring strategy with clear playbooks, measurable impact, and iterative improvements.

Software integration skills · 10%
Models must be integrated into services or workflows reliably using engineering practices. Evaluated in Round 3.

Weak: Can’t describe integration details, interfaces, or reliability considerations.
Meets the bar: Explains how inference/training artifacts are integrated and validated with basic reliability practices.
Strong: Demonstrates thoughtful integration design including versioning, fallbacks, latency/performance considerations, and testing.

Python and ML tooling · 8%
Effective hands-on development in Python is required for modeling and experimentation. Evaluated in Round 3.

Weak: Describes tooling at a high level without concrete implementation experience.
Meets the bar: Cites practical use of Python and ML/data libraries to run experiments and manage data transformations.
Strong: Shows depth in building maintainable code for experiments and production workflows (e.g., data pipelines, feature engineering patterns).

Ownership and communication · 6%
Mid-level independence requires clear communication of trade-offs and progress. Evaluated in Round 3.

Weak: Vague on ownership boundaries and cannot clearly communicate decisions or risks.
Meets the bar: Explains goals, decisions, and outcomes with clear documentation and stakeholder alignment.
Strong: Proactively identifies risks, communicates trade-offs, and drives work to completion with minimal prompting.

Interview questions

Questions for every interview round

Each round has an owner, a time box and questions that test the competencies above.

Round 1: Modeling and pipeline foundation
ML Engineering Lead · 50 min

Walk through a production ML project you shipped independently. What was the objective, what model did you choose, and how did you validate it before rollout?
Explain your approach to an end-to-end ML pipeline (data → features → training → artifacts). How do you ensure reproducibility and consistent training data?
Describe a time you needed to debug an ML training failure or data issue. What did you check first and how did you prevent recurrence?

Round 2: Evaluation and operational reliability
Staff Data Scientist · 55 min

How do you choose the right offline metrics for a modeling objective? Give an example where you changed the metric or sampling strategy based on results.
Describe a structured error analysis you performed (e.g., slicing, segmentation, calibration, or outlier analysis). What insights did it produce and what model change followed?
Tell us about monitoring you implemented for model quality and drift. What signals did you track, how did you set alerts, and what actions did you take when alerts fired?

Round 3: Integration and independent execution
Software Engineering Manager · 50 min

Describe how you integrated a model into a production workflow (batch or service). How did you handle versioning, compatibility, and validation of outputs?
Give an example of how you used Python and common ML/data libraries to build maintainable code for experiments or pipelines. What practices made it easier to iterate safely?
Share a case where you owned a well-defined scope end-to-end and coordinated with others. How did you communicate progress, risks, and trade-offs?

Red flags to watch for

Relies entirely on offline metrics and cannot explain production validation, monitoring, or degradation handling
Cannot describe how training data/features and model artifacts are reproduced or versioned
Uses vague evaluation practices with no clear metric rationale or error analysis methodology
Avoids discussing monitoring/alerting or has no example of investigating and fixing model issues in production

How much does it cost to hire a Machine Learning Engineer in Latin America?

In 2026, a senior Machine Learning Engineer in Latin America costs $8,250 to $9,500 a month all-in, versus $16,750 in the US: up to 51% less. Mid-level starts at $5,500.

Country

Mid-level

Senior

Argentina

$5,500

$8,250

Brazil

$6,125

$9,125

Colombia

$5,500

$8,250

Mexico

$6,375

$9,500

United States

$12,500

$16,750

Monthly cost in USD, all-in: the talent's salary plus Teilur's flat 20% fee, from our 2026 rates. The United States row shows what the same role typically costs there.

Compare costs in the calculator →

Frequently asked questions

What does a Machine Learning Engineer do?

Typical responsibilities include: develop and iterate ML models (e.g., ranking, classification, regression, or forecasting) based on product goals and data availability; build and productionize ML data pipelines and training workflows with reproducibility and monitoring in mind; integrate model outputs into downstream services or batch/streaming workflows used by product teams. The full list is in the template above.

How much does it cost to hire a Machine Learning Engineer in Latin America?

In 2026, a senior Machine Learning Engineer in Latin America costs $8,250 to $9,500 a month all-in, versus $16,750 in the US: up to 51% less. Mid-level starts at $5,500.

Can I customize this Machine Learning Engineer job description?

Yes. Teilur's free Job Description & Scorecard Generator rewrites it for your stack, seniority, time zone and hiring urgency in about 15 seconds. No sign-up needed.

Book a Free Consultation

Machine Learning Engineer Job Description Template