What I'm good at.

Spatial Computing & AR Engineering

This is where most of my production experience lives. I've shipped AR applications to tens of thousands of users across HoloLens and Android — not prototypes, production systems.

• HoloLens development with MRTK — spatial anchors, hand tracking, gaze interaction, shared coordinate systems for multi-device collaborative XR
• ARCore — Depth API, raw vs smoothed depth tradeoffs, ambient occlusion pipelines, plane detection, hit testing, UV coordinate transforms
• OpenGL ES and GLSL — custom depth shaders, background rendering, depth-based occlusion, UV texture coordinate systems
• Spatial computing constraints — field of view design, additive display optimization, BLE thin client architecture for wearables
• Production shipping — factory floor training apps, collaborative XR design tools, AR vehicle visualization at Mercedes-Benz R&D

The HoloLens factory floor training app and ARCore burst-view tools are representative examples. Currently building ARCore Depth API + Gemini Vision AI for spatially anchored object intelligence.

Android Engineering

I build modern Android apps that are reliable, maintainable, and pleasant to use. I'm comfortable with:

• Jetpack Compose for UI — including composing it over GLSurfaceView for AR applications
• Dependency injection with Hilt
• Multi-module app structures and clean architecture
• Performance profiling with Perfetto and StrictMode — ANR reduction, cold start optimization, main-thread I/O elimination
• Real-time data pipelines — live telemetry ingestion, dynamic map clustering, offline caching with retry/backoff
• Handling sensors, background behavior, and constrained hardware

Truckonnect (live on Google Play under BharatBenz/Daimler), SenseMap, and Egg Timer are examples across different scales and problem domains.

AI + Physical World Interfaces

I'm specifically interested in AI that understands and augments the physical world — not just answering questions, but reasoning about what's visible and generating spatially meaningful outputs.

• Gemini Vision for physical object understanding — identification, spatial position estimation, hierarchical knowledge generation
• Multimodal AI pipelines — camera frame capture, VLM API calls, structured JSON parsing, AR overlay rendering
• Stateless conversation architecture — history maintained client-side, efficient serverless backend
• On-device vs cloud tradeoffs for real-time AI in constrained environments
• Prompt engineering for structured spatial outputs — UV coordinates, component hierarchies, visibility estimation

Animus (talk to any physical object via AI) and the ARCore + Gemini research project are the clearest expressions of this direction.

Developer Tools & System Architecture

I enjoy building tools that help developers understand and work with their systems more clearly.

• Turning code into a graph of relationships and visualizing project structure
• Using AI to assist without replacing developer judgment
• Engine-level systems — physics, collision, spatial bounding volumes in PrimeEngine (C++)
• Shader authoring tools — node-based editor for shader parameter editing with observability and action logging

Product Thinking

Across all my work, I try to think in terms of:

• What problem this actually solves — and whether solving it matters
• How clearly the user can understand what's going on
• How the system behaves over time, not just in a demo
• How easy it will be to extend and maintain
• Performance on real devices, not just flagships