diya gangwar | Infra AI Engineer
AI Startup
San Francisco, CAOn-siteFull-time$140k – $220k / year
About this role
Diya Gangwar
Software Engineer | Machine Learning & Infra | CS @ Harvey Mudd
San Francisco, California, United States (US)
500 connections • 2,334 followers
About
Software engineer focused on ML infrastructure, backend systems, and deploying AI in real-world environments. I’ve built LLM-powered tools, distributed training systems, and data pipelines across startups, research labs, and enterprise teams, always with an emphasis on reliability and scale. Drawn to problems where machine learning meets systems design.
Total Experience: 2 years and 7 months
Experience
Infra AI Engineer - Zipline AI(https://www.linkedin.com/company/ziplineai) (Current)
Jun 2026 - Present (1 month) in San Francisco Bay Area
Department: Engineering and Technical • Level: Specialist
Zipline AI is a Data Infrastructure and Analytics company. Zipline AI is a next generation data platform for AI/ML, working with a select group of companies to improve the way they ship ai/ml products. Zipline AI has 10-20 employees (+40% YoY). Headquartered in Union City, California, United States. Has $7M in total funding.
AI ML Infra Engineer (Team Lead) - Applied Computing Research Labs(https://www.linkedin.com/company/applied-computing-research-labs)
Aug 2025 - May 2026 (9 months) in Claremont, California, United States
- Built Kubernetes-based ML training arena (SimKube) for RL benchmarking of cluster management agents
- Engineered distributed training infrastructure (AWS EC2, S3) to scale large-scale experiments
- Designed evaluation pipelines and sandboxed frameworks for safe ML agent interaction with live clusters
Department: Engineering and Technical • Level: Head
Applied Computing Research Labs is a Technology, Information and Internet company. ACRL is a company that provides open-source research and development in distributed systems. They apply simulation analysis to compute platforms to eliminate bottlenecks, optimize costs, and improve resource utilization without