Software Engineer - Agentic AI

Comcast · Media · Philadelphia, PA

Software Engineer role focused on designing, building, and operating intelligent agents end-to-end for Comcast's chat experience. The role involves rewriting a production application into an agentic platform, integrating AI across the SDLC, and ensuring reliable, scalable, and observable agent-based solutions. Requires experience with agentic AI systems, ADK, Gemini Enterprise, prompt engineering, and LLM evaluation techniques.

What you'd actually do

  1. Design, build, and operate agentic systems that replace or augment human workflows
  2. Apply AI across the SDLC, including: Requirements and spec-driven development, System and agentic workflow design, AI-augmented coding and refactoring, AI-driven testing and defect remediation
  3. Build production-ready agent-based solutions with reliability, scalability, and observability
  4. Implement agent-to-agent, agent-to-UI, and human–AI orchestration
  5. Ensure ethical, compliant, and responsible AI use.

Skills

Required

  • Agentic AI systems in production
  • Agent Development Kit (ADK)
  • Python
  • Gemini Enterprise or similar conversational AI platforms
  • Spec-driven development with AI integrated into the development lifecycle
  • Prompt engineering
  • Evaluation, iteration, and logging for AI outputs
  • AI-assisted coding
  • AgentOps and production monitoring for agent-based systems
  • LLM-as-a-Judge or comparable evaluation techniques
  • Agent-to-agent, agent-to-UI, and human–AI collaboration patterns

Nice to have

  • Curiosity to explore emerging AI capabilities
  • Critical thinking to assess AI outputs
  • Adaptability in fast-moving, ambiguous environments

What the JD emphasized

  • 2+ years building agentic AI systems in production
  • Hands-on experience with Agent Development Kit (ADK) and Python
  • Experience with Gemini Enterprise, or similar conversational AI platforms
  • Strong prompt engineering skills, including evaluation, iteration, and logging
  • Experience with LLM-as-a-Judge or comparable evaluation techniques
  • Proven ability to design agent-to-agent, agent-to-UI, and human–AI collaboration patterns

Other signals

  • design and build intelligent agents end to end
  • rewrite an existing production application into an end-to-end agentic platform
  • treat AI as a core part of the system—embedded across the full SDLC to deliver production-grade outcomes