Senior Staff Engineer with 19+ years building distributed systems, ML infrastructure, and cloud-native platforms at Intel, VMware, Walmart, and Cisco. Currently building the next layer of intelligent infrastructure.
A purpose-built Machine Learning Operating System with kernel-level optimizations for ML workloads. Includes Axon, an open-source universal model installer supporting HuggingFace, PyTorch Hub, TensorFlow Hub, and ModelScope — covering 80%+ of the ML ecosystem. Two USPTO provisional patents filed for kernel-level ML optimizations and universal model lifecycle management.
A high-performance, drop-in replacement for etcd written in Rust. Full gRPC API compatibility with etcd v3, MVCC with revision history, Raft consensus with PreVote, and validated against Kubernetes v1.35 with 34/34 compliance tests passing. 1.6x faster writes, 2x less memory, 10x faster K8s deployments.
Visual intelligence engine for real-time merchandise detection and catalog matching. Uses a YOLO + CLIP pipeline for video stream analysis — detect objects, extract embeddings, and match against product catalogs using FAISS. Targeting retail and e-commerce with edge deployment via ONNX Runtime.
MCP server-based context management tool for AI-powered development environments. Lets developers record and recall context shortcuts within tools like Cursor IDE, making agentic AI workflows significantly more efficient.
Open-source visual technical interview preparation platform with interactive algorithm visualizations, system design patterns, Kubernetes guides, and a modular content system that allows community contributions via simple GitHub pull requests.
Local-first observability platform for AI agents. Focused on LLM monitoring, agent debugging, and trace visualization for agentic AI systems. Designed for developers who need visibility into their AI pipeline without sending data to third parties.
ML-specific kernel scheduling, tensor memory management, GPU orchestration, and hardware-aware optimization layers for a purpose-built ML operating system. Filed August 2025.
Standard Model Interface (SMI) for cross-repository model installation, ONNX conversion, runtime registration, and unified inference across HuggingFace, PyTorch Hub, TensorFlow Hub, and ModelScope.
College of Technology, G.B.P.U.A&T, Pantnagar (1999–2004)
Professional Certification
Open to distinguished engineering roles, founding engineer positions, and technical collaborations for 2026.