Large Language Model R&D Engineer
Core
Develop and optimize LLM, VLM, embedding, and audio multimodal models based on open-source foundation models to build GUI-recognition and automated-control agents for medical post-processing software and devices.
Role type
Senior IC machine-learning engineer (multimodal models & agents)
Builds
GUI-recognition and automated-control agents for medical post-processing software and devices
Domain
Medical technology / AI / Multimodal models
Deliverable
production ML models
Required skills
Transformer architectures, LLMs, VLMs, PyTorch, PEFT (LoRA/QLoRA), SFT, DPO, dataset building, automated evaluation, model distillation, pruning, quantization (PTQ/QAT), TensorRT, ONNX Runtime, Python, Linux
Preferred skills
Medical device AI, agent development, tool calling, orchestration frameworks (LangChain/LangGraph), embedded optimization, C++, inference SDKs, vector databases
Technologies
LLaVA, TensorRT, ONNX Runtime, LangChain, LangGraph, PEFT, LoRA, QLoRA, SFT, DPO
Responsibilities
Fine-tune multimodal models using PEFT methods; build multimodal datasets from software screenshots and device commands; establish automated evaluation for recognition accuracy and hallucination rates; apply distillation, pruning, and quantization to reduce model size; optimize inference through KV cache and operator fusion; integrate VLM and LLM capabilities into automated agent loops; support model verification and risk analysis.
Seniority
Senior, hands-on IC
