Ursamin · 3 weeks ago
Staff Engineer (ML/LLM Fine-tuning)
United States
Full-time
Remote
Lead/StaffUrsamin is building an intelligent Patient Intake AI that interacts with patients like a trained medical assistant. They are seeking a hands-on Staff Engineer to set up and fine-tune an open-source LLM for patient intake, ensuring it can handle conversations intelligently and efficiently.
HealthcareHealth Care
Responsibilities
Model Setup: Select and configure the optimal open-source LLM for this use case. Base model will be provided. Setup complete training and inference pipeline on Hugging Face
Data & Training Pipeline: Inject, clean, and structure the Patient Intake Datasets provided by us. Prepare data for instruction tuning and conversational fine-tuning. Train the model using efficient methods like LoRA / QLoRA / SFT
Fine-tuning & Logic Implementation: Fine-tune the model on provided custom instructions. Implement conversational memory and conditional logic so the model understands patient context and drives the conversation forward
Model Output - Core Intelligence To Achieve: The final model must be able to: Ask relevant patient intake questions through model's trained specialty, Ask intelligent, context-aware follow-up questions based on previous answers, Provide relevant multiple-choice options within questions to speed up responses
Evaluation & Handover: Evaluate for accuracy, relevance and hallucination control. Document training process, hyperparameters, and evaluation metrics. Deliver the final fine-tuned model with inference ready API on Hugging Face
Qualification
Required
Prior work with healthcare / medical conversational data required
Hugging Face Ecosystem (Transformers, Datasets, PEFT, TRL, Accelerate)
Python
PyTorch
LLM Fine-tuning (LoRA, QLoRA, Instruction Tuning, SFT)
Experience with Llama 3 / Mistral / Gemma / Similar Open Source Models
Benefits
Meaningful Stock Options
Company
Ursamin
Care Coordination Platform.
Founded in 2020
2-10 employees
Funding
Current Stage
Early StageLeadership Team
John Rinn
Chief Information Officer
Company data provided by crunchbase