Research Engineer, Post-Training & Vision-Language Models

Cohesity•Santa Clara, CA
•$255,600 - $284,000•Hybrid

About The Position

Cohesity’s Gaia team is building an agentic document processing product that uses vision-language models (VLMs) to interpret enterprise documents, extract structured data, understand layouts, tables, and charts, and answer questions about document content. The Research Engineer works at the intersection of research and production, owning model fine-tuning, evaluation, and data pipelines that improve model understanding of enterprise data. The role involves open-ended problems, significant compute, and work that directly contributes to a shipping product. The engineer will evaluate and adapt VLMs, fine-tune models using supervised and reinforcement learning methods, build RL training platforms and evaluation harnesses, develop synthetic-data pipelines, and improve post-training recipes. The role also works with agentic engineering teams so agents can interpret and act on visual inputs. Employees within a reasonable commute of a Cohesity core office work onsite two to three days per week; candidates outside the area should be willing to relocate before starting and have the legal right to work in the job location.

Requirements

  • At least 5 years of professional experience in machine learning engineering or research.
  • Hands-on experience training, fine-tuning, using reinforcement-learning techniques with, or evaluating large language models or vision-language models.
  • A track record of taking models from experimentation to production.
  • Proficiency in Python.
  • Experience with PyTorch or JAX, the Hugging Face ecosystem, supervised fine-tuning, and VLMs or LLMs.
  • Experience building evaluation harnesses and benchmarks for model quality.
  • Experience with synthetic data generation and data-curation pipelines.
  • Ability to work across research and engineering, remain productive amid ambiguity and shifting priorities, and communicate results and tradeoffs clearly.

Nice To Haves

  • Master’s degree or PhD in Computer Science, Machine Learning, or a related field.
  • Experience with document AI, including OCR, layout analysis, and table and chart extraction.
  • Experience with visual grounding and region-level reasoning.
  • Experience with post-training and distillation for smaller or on-premises-deployable models.
  • Publications or open-source contributions in post-training, alignment, or multimodal models.
  • Prior experience in enterprise data protection, security, or infrastructure software.

Responsibilities

  • Evaluate, select, and adapt open-weight and API-based VLMs for document parsing, layout understanding, table and chart extraction, OCR-heavy content, and document question answering.
  • Fine-tune vision-language models using supervised fine-tuning and reinforcement learning techniques, including RLHF, RLAIF, RLVR, GRPO, and PPO-style methods.
  • Build reinforcement-learning training platforms, design reward signals, and optimize model reasoning.
  • Build and maintain evaluation harnesses to measure model performance across capability and reliability dimensions.
  • Create synthetic-data-generation pipelines for training and evaluation.
  • Develop and refine post-training recipes that measurably improve production model quality.
  • Collaborate with the agentic engineering team to enable agents to perceive and act on visual inputs such as documents, screens, and diagrams.
  • Debug training instabilities, regressions, and unexpected model behavior, and establish best practices for reproducible training and experiment tracking.
  • Balance research exploration with engineering rigor and reliable delivery.
  • Communicate experimental results and tradeoffs to technical and non-technical stakeholders.
  • Write and review production-quality code and run experiments regularly.

Benefits

  • Health and wellness benefits
  • Vacation
  • Paid holidays and refresh days
  • 401(k) retirement plan
  • Life and disability insurance coverages
  • Potential bonus compensation and/or equity grants
© 2026 Teal Labs, Inc
Privacy PolicyTerms of Service