Staff Agentic ML Engineer - Photoshop

Adobe Adobe · Enterprise · San Jose, CA +5 · Remote

Staff ML Engineer at Adobe, focusing on building advanced AI platforms for creative workflows. The role involves designing and implementing reasoning systems, tool orchestration, and multimodal integrations using LLMs and VLMs. Key responsibilities include LLM post-training, agent system architecture, practical agent development, image generation/editing with multimodal AI, and building evaluation frameworks. Requires expertise in LLMs, fine-tuning, agentic systems, and evaluation frameworks, with a strong foundation in ML and software engineering.

What you'd actually do

  1. Lead the development of next-generation AI agents that improve and extend Adobe's product ecosystem
  2. Drive the architecture and implementation of robust agentic systems that combine reasoning, retrieval, tool use, and multimodal capabilities
  3. Own end-to-end LLM fine-tuning pipelines — data curation, training, RLHF/DPO alignment, and evaluation
  4. Design and maintain comprehensive evaluation frameworks to measure agent quality, reliability, and safety
  5. Implement state-of-the-art LLM techniques for specialized creative applications

Skills

Required

  • Master's or Ph.D. in Computer Science, Machine Learning, Data Science, or a related field
  • 8+ years of industry ML experience
  • LLMs
  • fine-tuning (SFT, RLHF, DPO, PEFT/LoRA)
  • agentic system development in production
  • evaluation frameworks for LLM and multi-agent systems
  • PyTorch
  • LangChain
  • LangGraph
  • MCP
  • Agent Development Kit (ADK)
  • data structures
  • algorithms
  • software engineering principles

Nice to have

  • multimodal models (vision-language, image generation/editing)

What the JD emphasized

  • Demonstrated expertise in LLMs, fine-tuning (SFT, RLHF, DPO, PEFT/LoRA), and agentic system development in production
  • Deep experience building evaluation frameworks for LLM and multi-agent systems (LLM-as-judge, automated benchmarks, red-teaming)
  • Proficiency in PyTorch and agentic frameworks: LangChain, LangGraph, MCP, Agent Development Kit (ADK)
  • Experience with multimodal models (vision-language, image generation/editing) is a strong plus

Other signals

  • LLM
  • VLM
  • agent system
  • fine-tuning
  • evaluation framework
  • multimodal