Lead AI Engineer (vision Model Customization, Vlm)

Capital One Capital One · Banking · New York, NY +3

Lead AI Engineer focused on customizing vision models (VLMs) and optimizing large-scale AI systems, including foundation model training and LLM inference. The role involves designing, developing, testing, deploying, and supporting AI software components, leveraging technologies like AWS, Huggingface, VectorDBs, and Nemo Guardrails. Emphasis is placed on improving performance (scalability, cost, latency, throughput) of production AI systems and contributing to the technical vision for foundational AI systems.

What you'd actually do

  1. Design, develop, test, deploy, and support AI software components including foundation model training, large language model inference, similarity search, guardrails, model evaluation, experimentation, governance, and observability, etc.
  2. Leverage a broad stack of Open Source and SaaS AI technologies such as AWS Ultraclusters, Huggingface, VectorDBs, Nemo Guardrails, PyTorch, and more.
  3. Invent and introduce state-of-the-art LLM optimization techniques to improve the performance — scalability, cost, latency, throughput — of large scale production AI systems.
  4. Contribute to the technical vision and the long term roadmap of foundational AI systems at Capital One.

Skills

Required

  • Python
  • Go
  • Scala
  • Java
  • Computer Science
  • AI
  • Electrical Engineering
  • Computer Engineering

Nice to have

  • AWS
  • Google Cloud
  • Azure
  • Huggingface
  • VectorDBs
  • Nemo Guardrails
  • PyTorch
  • LLM Inference
  • Similarity Search
  • Guardrails
  • Memory
  • C++
  • C#
  • Golang
  • training optimization
  • inference optimization
  • hardware utilization
  • latency
  • throughput
  • cost optimization

What the JD emphasized

  • Vision model customization
  • VLM

Other signals

  • customization of vision models
  • large scale production AI systems
  • foundation model training
  • large language model inference
  • similarity search
  • guardrails
  • model evaluation
  • experimentation
  • governance
  • observability