Healthcare AI Engineer - Build Trusted Models for Fast Care

Tala Health; San Francisco (CA); Hybrid; USD 140,000 - 190,000; Full time

Job summary

Tala Health in San Francisco, CA is hiring a Machine Learning Engineer to turn clinical data into world‑class AI that clinicians can trust from day one. You’ll fine‑tune foundation models and ship models that support the full patient journey, balancing innovation with privacy and safety requirements.

You'll use Python, PyTorch/TensorFlow, and HuggingFace, building scalable training pipelines and contributing to model‑serving integrations (RAG, LoRA, quantization) for low‑latency inference in

Qualifications

• 4+ years of ML engineering or research with deep learning.
• Proficiency in Python, PyTorch or TensorFlow.
• Experience with distributed training and model evaluation.
• Familiarity with healthcare data standards (HIPAA) is a plus.
• M.S. or Ph.D. in CS/EE/Stats or related field preferred.

Responsibilities

• Fine-tune and retrain large language and multimodal models on de-identified clinical data sets.
• Build scalable training pipelines (data prep, augmentation, distributed jobs).
• Design automated evaluation frameworks to track accuracy, bias, and safety metrics.
• Run experiments and hyperparameter sweeps; document insights.
• Collaborate with Data, Product and Clinical teams to iterate rapidly.
• Contribute to model-serving integrations (RAG, LoRA, quantization) for low-latency inference.

Skills

Python

PyTorch

TensorFlow

HuggingFace

Distributed training

Model fine-tuning

Research experience

GPU/TPU orchestration

Education

MS/PhD in CS/EE/Stats or related

Tools

DDP

DeepSpeed

RAG

LoRA

Quantization

HuggingFace

Job description

Tala Health in San Francisco, CA is hiring a Machine Learning Engineer to turn clinical data into world‑class AI that clinicians can trust from day one. You’ll fine‑tune foundation models and ship models that support the full patient journey, balancing innovation with privacy and safety requirements.

You'll use Python, PyTorch/TensorFlow, and HuggingFace, building scalable training pipelines and contributing to model‑serving integrations (RAG, LoRA, quantization) for low‑latency inference in