Research Engineer, Post-Training & Vision-Language Models
Core
Research Engineer building agentic document processing products using vision-language models (VLMs) to read enterprise documents as images, extract structured data, and answer questions.
Role type
Senior Research Engineer (Post-Training & VLMs)
Builds
Agentic document processing product for enterprise data
Domain
AI-powered data security and management
Deliverable
production ML models
Required skills
Python, PyTorch, Hugging Face, supervised fine-tuning, VLMs/LLMs, building eval harnesses, synthetic data generation
Preferred skills
Document AI (OCR, layout analysis), visual grounding, post-training/distillation, publications in post-training/alignment
Responsibilities
Evaluate and adapt VLMs for document parsing and layout understanding; Fine-tune VLMs using SFT and RL techniques (RLHF, GRPO, PPO); Build RL training platforms and reward signals; Develop evaluation harnesses; Create synthetic data pipelines; Debug training instabilities; Collaborate with engineering teams on agentic perception
Seniority
Senior, hands-on IC