Machine Learning Engineer

HMH

19d • $110,000 - $115,000 • Remote

About The Position

NWEA, a division of HMH, develops pre-K–12 assessments and education services. HMH is a learning technology company serving more than 50 million students and 4 million educators in 150 countries. This role builds and operates the models inside HMH Pulse, which use assessment results to provide teachers with explainable information about student performance, readiness, and expected growth. Models cover data representing millions of students and produce outputs used in classrooms.

The Machine Learning Engineer works across the full model lifecycle on AWS SageMaker using Python, including data exploration, training, validation, release, serving, monitoring, and iteration. The role suits either a strong modeller who wants to run models in production or a strong ML engineer who wants to work on inference and statistical problems. The engineer joins a stream-aligned team, works alongside a senior or staff engineer, and is expected to know their models’ behavior in production in depth.

Work includes exploring and checking data; training and validating models with SageMaker and MLflow; registering, promoting, and deploying traceable model versions through code review and CI; serving real-time and batch predictions using a feature store and Snowflake; monitoring prediction quality, failures, and drift; improving models based on production evidence; contributing analysis and evidence to design documents (RFCs); and supporting analysts, product, and data teams in understanding and using model outputs.

The team’s problems include extending models across subjects or curricula without code forks, rescoring large student populations while live services continue operating, running model versions in parallel and retiring versions safely, detecting data and model drift, and improving training and serving efficiency without changing model outputs. The team emphasizes evidence-based decisions, thorough written analysis and operational documentation, production-quality model code, accountable model ownership, and agent-assisted engineering with careful review.

The position is remote and may be performed from any U.S. state. It is an experienced individual-contributor role at level IC4, working alongside a senior or staff engineer. The stated salary range is $110,000–$115,000; compensation depends on factors including skills, experience, education, and location. The application window is anticipated to close October 23, 2026.

Requirements

• A track record of shipping models or machine-learning systems that others depended on.
• For IC4, typically three or more years of hands-on work, or a doctorate combined with relevant practical experience.
• A master’s or Ph.D. in a quantitative field; psychometrics is included.
• Idiomatic, typed, tested Python that passes code review, along with comfort with CI/CD and Terraform-managed infrastructure.
• Solid SQL skills.
• A usable statistics foundation, including the ability to explain posterior distributions, likelihoods, and calibration plots to colleagues.
• Disciplined exploratory data analysis, including identifying unusual data and explaining its significance.
• Excellent writing skills, including the ability to clearly document analyses, designs, results, and incidents in actionable form.
• Working knowledge of AWS, ideally SageMaker, and an experiment tracker such as MLflow.
• Confidence to identify when a model is not ready to ship and provide evidence supporting that judgment.

Nice To Haves

• Psychometrics or education measurement.
• Cognitive diagnostic models or Bayesian networks.
• Probabilistic programming, such as Stan or PyMC.
• R.
• Snowflake and dbt.
• Terraform.
• GitHub Actions.
• Knowledge graphs or ontologies.
• Experience with K–12 assessment data.

Responsibilities

• Explore and inspect data before and after model fitting, assess how models perform on unseen data, and document findings for review.
• Train and validate models using Python and SageMaker training jobs and pipelines, track experiments in MLflow, and validate against held-out cohorts before promotion.
• Register, promote, and deploy model versions with defensible lineage covering data, code, weights, and deployment destination; release changes through code review and CI.
• Operate real-time endpoints and batch scoring backed by a feature store and Snowflake, and understand and tune latency, cost, and capacity.
• Monitor prediction quality, failure rates, and drift using dashboards and alerts; participate in monthly production-health reviews, conduct blameless postmortems, and implement follow-up fixes.
• Use production evidence to propose model improvements and contribute analysis and evidence to team design documents (RFCs).
• Help analyst, product, and data teams understand model outputs and use them effectively.
• Develop or improve models and systems to support multiple subjects or curricula without code forks, large-scale rescoring, side-by-side model versions, drift detection, and efficient training and serving.

What This Job Offers

• Job Type: Full-time
• Career Level: Mid Level