Software engineer · AI systems researcher

I build systems that are useful, secure, and ready to scale.

I’m Kshitij, a UW Graduate student and Software Engineer in Seattle. I've spent five years shipping software at Microsoft and now build AI agents and evaluation systems at UW’s eScience Institute.

Kshitij Dani standing in front of cherry blossoms
60Musers reached monthly
90%infrastructure cost reduction
500K+premium users served
5+ yrsshipping production systems

What I bring

From model behavior to production reliability.

My work sits where AI meets systems engineering: turning promising capabilities into software that can be evaluated, operated, secured, and trusted.

AI systems & agents

LLM agents, RAG, evaluation systems, embeddings, synthetic personas, and reliable workflow integration.

  • Python
  • FastAPI
  • Pydantic
  • LiteLLM
  • LangGraph
  • RAG

Backend & distributed systems

High-scale services, service boundaries, data integrations, observability, and operational reliability.

  • C#
  • .NET
  • Django
  • SQL
  • REST APIs
  • Distributed systems

Cloud, security & delivery

Cloud-native deployments, CI/CD, authentication, access tokens, certificates, and secure system design.

  • Azure
  • Kubernetes
  • CI/CD
  • Cosmos DB
  • Azure SQL
  • Security

Experience

Built across research and industry.

AI research at UW, enterprise systems at Microsoft, and conversational AI at Avaamo.

Full experience

University of Washington eScience Institute · SSEC

Graduate Researcher — AI Systems

Building LLM-powered systems for research engineering collaboration and evaluation infrastructure for sensitive conversational agents.

Microsoft · Viva Insights

Software Engineer II

Built enterprise analytics platforms and secure backend infrastructure for Copilot and Viva Insights.

Microsoft · Viva Insights

Software Engineer

Led high-scale email experiences and backend services for hybrid-work products.

Microsoft · MileIQ

Software Engineer

Owned a production reporting service and its cloud delivery infrastructure.