Computer Vision Services
Dockerised CV microservices built on PyTorch, TensorFlow, and PaddlePaddle — covering classification, segmentation, and keypoint detection, ready to drop into any pipeline.
View on GitHub ↗Portfolio / engineer / writer
I work at the intersection of GenAI, Computer Vision, and Deep Learning. I love building sophisticated products.
Selected work
Projects
Nirikshan
End-to-end video analytics pipeline powered by AI and deep learning. Handles real-time scene understanding, object tracking, and event-driven alerts at scale.
View on GitHub ↗Multi-Agent Protocols Survey
Exploration of modern multi-agent protocols — MCP, ACP, and A2A — with comparisons of architectures and real-world applicability.
View on GitHub ↗OpenCV Snippets
A curated collection of practical Python recipes for image processing with OpenCV — filtering, transforms, contour analysis, and feature extraction.
View on GitHub ↗Local LLM Quant Bench
A local benchmark harness for comparing LLM quantization/runtime variants under fixed prompts and decoding settings.
View on GitHub ↗HeyBagh
Content-based image retrieval engine using ResNet-18 embeddings and Milvus vector database — find and discover visually similar images via a Gradio interface.
View on GitHub ↗Recent writing
Blogs
What is Tokenization & Why LLMs Don't See Words
Read post ↗Workflow and Agent Patterns for Intelligent Systems
Read post ↗CNN Series: ResNet
Read post ↗CNN Series: VGGNet
Read post ↗CNN Series: AlexNet
Read post ↗LearningByDoing Series: Building Blocks of Neural Network - Activation Functions
Read post ↗Reach out