Sainath Ganesh

Magic Leap – Google Partnership

• Shipped a productionized XR pipeline for the Google Capture App, integrating with backend services to facilitate 3D Gaussian Splats & NeRFs. Launched at 500+ locations around SF & NYC.
• Built a Gemini-powered 3D calibration tool for Google Maps XR, accelerating QA validation by 10x.
• Led end-to-end Out of Box Experience for the Magic Leap 2, cross-functionally with UX & Product

Healiam - Preventive HealthTech Startup (Link)

• Co-founded Healiam, a preventive blood-testing startup breaking through the barriers of insurance making proactive healthcare accessible.
• Went 0-to-1, setting up the technology, architecture and scaling the business to $100K+ ARR.
• Architected the Full stack end-to-end — infrastructure, backend services, cross-platform frontend, AI initiatives, deployment, CI/CD, and HIPAA compliance.

University of Illinois, Urbana-Champaign

• Optimized upstream video pipelines using Gaze-tracked Foveation, reducing bandwidth by 70% and enabling improved ML inferencing quality on edge and constrained networks.
• Designed a real-time rendering pipeline for ILLIXR 3D mesh reconstruction, streaming meshes over websockets, optimizing serialization and transfer overheads with protobuf.

Shell India Markets, Bangalore

University of Illinois, Urbana-Champaign — GPA 4.0

Vellore Institute of Technology — GPA 3.6

Consistent top finishes in several national and international hackathons.

Agentic AI Summit - University of California, Berkeley

Speaker at Agentic AI Summit, UC Berkeley after winning LLM Agents hackathon among 1100+ universities and 800+ companies worldwide(Link)

Toycathon 2021 — Ministry of Education, India

Recognized and commended for innovative projects leveraging VR + AI(Link)

COVID AR Hackathon — NVIDIA, Indian Ministry of Commerce and Trade

Winner of UC Berkeley LLM Agents Hackathon

Orchestrated a multi-agent system via LangChain and GPT Realtime (via Twilio) that automates conversations with moving companies and negotiates the best quotes, saving users $500–$2,000 per move and reclaiming 3–5 hours of manual effort(Link)

Improved facial encoding search speed by 15x, integrating classical computer-vision filtering via KD-Trees as a preprocessing step to deep convolutional neural networks, reducing compute and improving end-to-end processing by 250%.

Designed and implemented lane-keep assist, pre-collision warning and avoidance, adaptive cruise control and other advanced driver-assistance features, enabling vehicle autonomy with cameras + ML fusion(Link)

IVCCPS’20, LNEE Springer

Published research on applying evolutionary algorithms to optimize neural networks for self-driving applications in simulated environments (Paper)