AI Engineer

Discover International; New York (NY); Sur place; USD 120 000 - 180 000; Plein temps

Résumé du poste

Discover International is seeking an ML engineer in New York to advance healthcare AI products. You will adapt and fine-tune LLMs on healthcare data, build retrieval-augmented systems, and design autonomous AI workflows that plan, call tools, and reason through multi-step reviews.

You will work with leadership to set metrics, validate improvements, and ensure secure, cost-effective deployment on sensitive patient information.

Qualifications

• 3+ years building ML or AI systems that real users depend on.
• Production experience with LLM-based agents and retrieval-augmented generation.
• Hands-on fine-tuning with LoRA or similar parameter-efficient methods in PyTorch / Hugging Face.
• Real-world use of RL or preference optimization on language models.
• Excellent Python and engineering discipline, plus a rigorous approach to evaluation.

Responsabilités

• Making models smarter: adapt and fine-tune large language models on specialized healthcare data, and own the training data, runs, and versioning
• Improving learning from experts: turn feedback from in-house medical reviewers into training signal using preference-based methods (DPO, RLHF, reward models)
• Grounding answers in evidence: build retrieval systems so every AI conclusion can be traced back to the document it came from
• Building autonomous workflows: design AI agents that plan, call tools, and reason through multi-step reviews
• Proving it works: set up the metrics and test suites that show whether a change actually helped
• Shipping it: work with engineering and infrastructure teammates to run everything securely, quickly, and cost-effectively on sensitive patient data

Connaissances

ML systems

LLM agents

Python

RLHF experience

Outils

PyTorch

Hugging Face

Description du poste

This well-funded, growth-stage healthcare technology company pairs healthcare expertise with AI to remove waste from the U.S. healthcare system. Its generative AI is already live, helping insurers find costly errors in medical payments. The next stage is making that AI more accurate, more explainable, and more reliable, and they want an engineer who has done it before. The team is lean, decisions move fast, and engineers work directly with leadership.

Responsibilities

• Making models smarter: adapt and fine-tune large language models on specialized healthcare data, and own the training data, runs, and versioning
• Improving learning from experts: turn feedback from in-house medical reviewers into training signal using preference-based methods (DPO, RLHF, reward models)
• Grounding answers in evidence: build retrieval systems so every AI conclusion can be traced back to the document it came from
• Building autonomous workflows: design AI agents that plan, call tools, and reason through multi-step reviews
• Proving it works: set up the metrics and test suites that show whether a change actually helped
• Shipping it: work with engineering and infrastructure teammates to run everything securely, quickly, and cost-effectively on sensitive patient data

Qualifications

• 3+ years building ML or AI systems that real users depend on
• Production experience with LLM-based agents and retrieval-augmented generation
• Hands-on fine-tuning (LoRA or similar parameter-efficient methods) in PyTorch / Hugging Face
• Real-world use of RL or preference optimization on language modelsExcellent Python and engineering discipline, plus a rigorous approach to evaluation

Nice to Haves

• GPU and distributed training experience
• Experience in healthcare or another regulated industry
• Familiarity with common LLM orchestration and experiment-tracking tools