Atul Vaish
Principal Architect | Engineering Leader
Personal technical site documenting independent applied engineering work in GPU optimization, edge AI inference, secure systems, and performance-critical platforms across x86 and ARM environments.
Hardware Platforms Used in Independent Prototyping & Research
Personal hardware testbed used for hands-on prototyping, benchmarking, and system-level experimentation.
Selected Technical Prototypes & Experiments
Independent technical work (2025–26) focused on GPU acceleration, system-level AI engineering, and performance benchmarking.
1. GPU Acceleration for Visual Search
High-performance visual search and object detection pipelines optimized using CUDA.
View on GitHub →
2. Agentic Compute on GPU
Parallelizing AI agent workflows and decision-making loops on GPUs.
GitHub / PyPI →
3. SentinelChange AI
Deep learning models for high-accuracy change detection in satellite imagery.
View on GitHub →
4. CUDA Softmax Benchmark
Benchmarking softmax kernel implementations for LLM inference.
View on GitHub →
5. CUDA GEMM Benchmark
Performance analysis of GEMM operations across GPU architectures.
View on GitHub →
6. Autonomous Local Research Agent
Local agent for technical research automation and document synthesis.
View on GitHub →
8. Voice LLM Deployment on Jetson Orin
End-to-end voice-to-LLM inference on constrained devices.
GitHub Repo →
10. Urban Digital Twin (Unity / Cesium)
3D Tiles integration and simulation for large-scale urban models.
Watch Demo →Cryptography & CloudHSM — Prior Training Work
Sample recording from hands-on cryptography and AWS CloudHSM training delivered as part of prior professional engagements.
Technical Blogs & Insights
Deep-dive write-ups and implementation notes from applied systems work.