Applied Scientist - LLMs

Semantic Health is on a mission to improve care delivery and operational inefficiencies by transforming the use of unstructured data in healthcare's revenue cycle. Our machine learning powered medical coding and auditing platform uses cutting edge deep learning to streamline manual and error-prone medical coding and auditing processes in health organizations. We help health organizations improve data quality, optimize reimbursements, and enable real-time access to actionable data for use across the health system.
At Semantic Health, we combine the clinical and business expertise of doctors and successful entrepreneurs, with the technical skillset of top ML researchers. We are backed by leading institutional investors who have driven companies our size to multi- billion dollar valuations.
We’re seeking a motivated and driven scientist to work on our foundation language models. We value scrappy and product-focused thinkers that thrive by solving complex and challenging problems.
As an Applied Scientist - LLM, you will:

• Implement, train, fine-tune, and optimize LLMs for specific applications in healthcare, ensuring their accuracy and efficiency.
• Utilize Kubernetes and containerization for the deployment and scaling of LLMs across multiple GPUs and servers.
• Oversee the process of data preprocessing, ensuring the quality and relevancy of data fed into the LLMs.
• Apply methods such as Retrieval-Augmented Generation (RAG) and Direct Preference Optimization (DPO) for enhancing model outputs.
• Stay current with the latest advancements in open-source LLM technologies.
• Work closely with ML engineers, software engineers, and healthcare professionals to integrate LLM solutions into practical applications.

• Applying and refining large language models, preferably with 2+ years in the field.
• Kubernetes, container technologies, and multi-GPU deployment strategies.
• Python, with experience in frameworks like PyTorch or TensorFlow. Familiarity with CUDA for GPU acceleration.
• Data cleaning, preprocessing, and manipulation, with at least 2 years of work in this area.
• Fine-tuning LLMs for specific tasks or industries, particularly in a Q&A context.
• LLM frameworks and models like LangChain, OpenLLM, vLLM, and Llama2.

Bonus points if you have experience with:

• Prior work with healthcare datasets, understanding of medical terminologies and compliance (e.g., HIPAA).