Senior Software Engineer, Ai/ml, Geo and Gemini App

Google Google · Big Tech · New York, NY +1

Senior Software Engineer role focused on building AI/ML features for Google Geo products, involving agentic workflows, evaluation frameworks, and serving infrastructure enhancement. Requires experience with LLMs, RAG, and model quality optimization.

What you'd actually do

  1. Build context engineering pipelines, agentic workflows with tool usage, and robust evaluation frameworks.
  2. Analyze model behavior, creating high-quality evaluation datasets to identify weaknesses and guide performance improvements.
  3. Enhance serving and evaluation infrastructure to power new user-facing features.
  4. Act as a technical bridge between engineering, product, UX, and research teams to translate user needs into valuable features.
  5. Analyze user metrics and model outputs to enhance personalization and overall system helpfulness.

Skills

Required

  • C++
  • backend infrastructure
  • distributed systems
  • software testing
  • software maintenance
  • software launching
  • software design
  • software architecture
  • Large Language Models (LLMs)
  • Generative AI
  • prompt engineering
  • retrieval-augmented generation (RAG)
  • model quality optimization
  • loss analysis
  • quality hillclimbing

Nice to have

  • data structures
  • algorithms
  • technical leadership
  • fast-paced environments
  • startup-like environments
  • data analysis
  • experimental design
  • statistical methods
  • TensorFlow
  • JAX
  • PyTorch
  • user metrics logging
  • A/B testing

What the JD emphasized

  • 5 years of experience developing C++ backend infrastructure components within distributed systems.
  • 3 years of experience testing, maintaining, or launching software products, and 1 year of experience with software design and architecture.
  • Experience with Large Language Models (LLMs) and Generative AI concepts (e.g., prompt engineering, retrieval-augmented generation (RAG)).
  • Experience with model quality optimization using methods such as loss analysis or quality hillclimbing.

Other signals

  • building agentic workflows
  • evaluating model behavior
  • enhancing serving infrastructure