Machine Learning Engineer, Evaluation
At a glance
Mid-level Machine Learning Engineering role at mpathic2. Remote, United States · full-time · no salary stated · we estimate $142,500–$230,000 across the United States.
mpathic2 has not published a salary for this role. The figure below is our estimate, not theirs.
aimlroles summary, based on the employer's posting.
What you'll do
- Build and support deployed ML models and services alongside engineering and product teams
- Set up evaluations that reflect product goals and the conditions systems face in practice
- Create datasets, benchmarks, and automated regression checks for safety, empathy, and risk
- Choose metrics, compare results across relevant data groups, and investigate model failures
- Run experiments, connect offline scores to customer-facing performance, and share findings with clinicians, product leads, and customers
What you bring
- You have shipped ML systems through practical work, usually for two to five years or an equivalent background
- You bring strong Python ability, familiarity with standard ML frameworks, and sound foundations in model training and validation
- You have worked with evaluation metrics, benchmark creation, and structured analysis of errors
- You understand statistical inference and experiment design, and can spot leakage, evaluation bias, and missing test coverage
Who this fits
This mid-level role suits a hands-on, generalist ML engineer who wants to focus on evaluating model quality and reliability while contributing to production systems. It is remote within the United States.
From the employer
About mpathic
mpathic builds AI that understands clinical and human conversations: analyzing therapy and clinical-trial sessions, producing expert-labeled data for AI builders, and publishing open safety benchmarks.
About the role
We’re looking for a Machine Learning Engineer to build and improve ML-powered products, with a particular focus on evaluating their quality and reliability. You’ll work across the ML lifecycle, combining hands-on engineering with rigorous experimentation to understand how systems perform and where they need to improve.
This is a generalist ML engineering role with an evaluation focus. You’ll contribute to production systems while helping the team make sound decisions about models, data, and product changes.
What you’ll do
- Build, deploy, and maintain ML models and services in collaboration with engineering and product teams.
- Design evaluation approaches that reflect product needs and real-world operating conditions.
- Create and maintain evaluation datasets, benchmarks, and automated regression tests that measure safety, empathy, and risk.
- Select appropriate metrics, analyze performance across meaningful data segments, and investigate failure modes.
- Design experiments and use statistical methods to judge whether an improvement is real.
- Connect offline evaluation results with what customers see in production.
- Develop tools that make evaluation repeatable and useful throughout development and release workflows.
- Communicate findings, limitations, and recommendations clearly to clinicians, product leads, and customers.
What you’ll bring
- Practical experience developing and shipping ML systems, typically 2–5 years of relevant work or equivalent.
- Strong Python skills and familiarity with common ML frameworks.
- Solid ML fundamentals, including training, validation, generalization, and common sources of model error.
- Experience with evaluation metrics, benchmark design, and systematic error analysis.
- An understanding of statistical inference, sampling, confidence intervals, and experimental design.
- Experience identifying data leakage, evaluation bias, and gaps in test coverage.
- Good software engineering practices, including testing, version control, and code review.
- Skill at turning ambiguous product goals into measurable quality criteria.
Helpful additional experience
- A/B testing and production performance monitoring.
- Human evaluation, annotation guidelines, and inter-rater agreement.
- Speech or conversational data: transcripts, diarization and multi-speaker setups.
- Evaluating ranking, retrieval, recommendation, or generative AI systems.
- Robustness testing, calibration, fairness analysis, or distribution shift detection.
- Building evaluation tooling and integrating quality checks into CI/CD workflows.
- Clear, direct communication, both written and verbal.
- Startups and cross-functional teams.