Senior AI Engineer (REMOTE)

EnableComp•Franklin, TN

2d•Remote

About The Position

The Senior AI Engineer is a hands-on individual contributor who builds, trains, and evaluates machine learning models. The role combines knowledge engineering and data analysis with applied reinforcement learning, and owns work end to end, from data and experiment design through training, evaluation, and reporting. The role works closely with domain experts and communicates results to technical and non-technical audiences. Roughly half of the role focuses on reinforcement learning and model training, with the remainder split between knowledge engineering, data analysis, and evaluation. The engineer is responsible for developing approaches that adapt as underlying knowledge evolves, while complying with HIPAA and EnableComp data policies and using de-identified or synthetic data wherever possible.

Requirements

Bachelor’s degree in computer science, machine learning, statistics, or a related field; equivalent hands-on experience will be considered.
At least 3 years of hands-on experience training or fine-tuning language models.
Demonstrated reinforcement learning experience, including designing an environment and reward function end to end and training a policy in it.
Working knowledge of policy optimization methods for language models, such as PPO, DPO, or GRPO, including their failure modes.
Strong Python and PyTorch skills, with fluency in the Hugging Face ecosystem and parameter-efficient fine-tuning techniques.
Experience building and evaluating LLM-based agents or multi-turn systems.
Experience designing evaluations for language models, including held-out sets and automated grading.
Ability to analyze complex, detailed documentation and translate it into precise, structured, testable form.
Clear technical writing and ability to explain experiments and results concisely.
Comfort working directly with domain experts who are not machine learning practitioners.
Ability to perform each essential duty satisfactorily, with reasonable accommodation available for qualified individuals with disabilities.

Nice To Haves

MS or PhD degree.
Experience working with healthcare data or in another highly regulated industry.
Experience training models to produce structured or schema-constrained output.
Experience with a broad range of reinforcement learning algorithms and advanced reward design.
Experience with knowledge representation approaches such as knowledge graphs or ontologies.
Experience generating and validating synthetic data for model training.
Publications or open-source contributions in reinforcement learning, language model fine-tuning, or model evaluation.
Experience handling sensitive data, including protected health information (PHI) under HIPAA.
Equivalent combination of education and experience will be considered.

Responsibilities

Work with domain experts to capture specialized knowledge and represent it in structured, usable form.
Analyze source material and datasets for quality, consistency, and coverage, and resolve issues with appropriate stakeholders.
Perform exploratory and statistical analysis to inform data, modeling, and evaluation decisions.
Create and curate training and evaluation datasets using automated and human validation.
Design reinforcement learning environments, reward functions, and verifiers.
Train and fine-tune language models using supervised and reinforcement learning methods such as PPO, DPO, and GRPO.
Establish baselines, run ablations, and diagnose issues such as reward hacking, instability, and data leakage.
Follow research literature and bring promising techniques into practice.
Manage experiments rigorously, including versioned configurations, reproducible runs, experiment tracking, and compute budgeting.
Design and extend evaluation frameworks, metrics, and benchmarks.
Compare models and approaches through structured experiments and A/B testing.
Document results, failures, and next steps in clear technical write-ups.
Present findings to engineers, domain experts, and business leaders in actionable language.
Design data, training, and evaluation approaches that can adapt as underlying knowledge evolves.
Comply with HIPAA and EnableComp data policies, using de-identified or synthetic data wherever possible.
Maintain timely and regular attendance.
Practice and adhere to EnableComp’s Core Values, Vision, and Mission.