Senior ML Engineer

Truemed enables health and wellness merchants to accept $150+ billion in HSA/FSA funds. At the center of this is determining whether products qualify for pretax spending—a classification problem with regulatory, financial, and clinical consequences at scale.

We solve this using cutting-edge ML technologies that enable us to scale the number of merchants and product categories we can support while maintaining industry-leading accuracy requirements. As a Senior ML Engineer, you will run the experiments that decide how that system improves, and ship the winners to production.

What you'll do

• Run the experiments that move classification accuracy: form a hypothesis, test it against labeled data, and ship what wins

• Design and iterate on classifiers that evaluate products against eligibility rules, across prompting, retrieval, fine-tuning, and multi-model approaches

• Dig into misclassifications to find failure modes, and turn each one into the next experiment

• Build and extend the evaluation sets and metrics that tell us when a model is better, including how we label ground truth

• Track new models and research, and test the ones that could change our accuracy or cost curve

• Extend classification coverage to new product categories as Truemed expands

• Find new applications for our models and data beyond product classification, and prototype the ones worth pursuing

• Collaborate with clinical ops and legal teams to turn eligibility standards into labeled data and testable system behavior

• Write up results so the team makes model decisions from evidence, and help raise AI fluency across engineering

What you bring

• 5+ years of engineering or applied research experience, with at least 2 years of hands-on work with LLMs or other ML models on real problems

• A track record of rigorous experiments: clear hypotheses, sound baselines, ablations, and conclusions you would defend

• You thrive and enjoy looking at data with your eyes for hours and consider it part of the workflow in any ML work

• Comfort designing evaluation methodology for AI systems—defining what ‘correct’ means and measuring it

• Hands-on experience with prompting, fine-tuning, retrieval, or agent design on foundation models (OpenAI, Anthropic, open-weight models)

• Fluency in Python and enough engineering depth to take an experiment to production

• Senior-level ownership: you've taken an ambiguous ML problem from first experiment to production on your own

• Comfort with ambiguity—the right approach isn't always known upfront and has to be discovered through experimentation

• Public evidence of how you run experiments: a paper, open-source ML work, or a technical write-up of results you would defend

• Bonus: experience with healthcare or regulatory classification, or multi-model orchestration