I design products that feel simple, even when they're not.
Currently architecting complex AI workflows and an autonomous multi-agent design & engineering system.
Know me better.
A deeper look into my role, engineering process, and what drives my work forward.
As Founder of VRSA Analytics, I own end-to-end product engineering and system design for e-commerce, B2B, and B2C platforms. I lead architecture, client delivery, and technical execution — shipping production systems (including inventory platforms and AI chatbots) while generating ₹3.5L+ in revenue. I also take on selective freelance work building multi-tenant SaaS products with strong focus on reliability, cost-efficient LLM usage, and clean CI/CD.
I start with deep requirement discovery, then design systems that stay simple on the surface even when the internals are complex. Most of my work revolves around multi-agent architectures (LangGraph supervisor patterns, Temporal durable workflows, MCP tool integrations, Redis checkpointing, and layered guardrails). I obsess over observability, fault tolerance, and measurable outcomes — then ship the whole thing end-to-end from architecture to production.
I love good food, short trips to new places, and spending hours experimenting with the latest open-source tools and AI frameworks. Whether it's trying a new cuisine, exploring a different city, or tinkering with the newest agent libraries just for fun — that's how I recharge.
Skills & Tech Stack.
Core technologies, AI agent frameworks, evaluation benchmarks, and battle-tested tools I use to build production-ready systems.
LangGraph
Stateful, cyclical multi-agent graph workflows & human-in-the-loop
Google ADK
Multi-modal agent orchestration, Gemini tool routing & function calling
MCP (Model Context Protocol)
Standardized tool & resource orchestration for autonomous agents
LangChain
Composable LLM application chains, memory systems & tool ecosystem
LlamaCloud
Managed data parsing, ingestion pipelines & production index retrieval
DeepEval
Unit testing for LLMs, hallucination benchmarking & RAG evaluation metrics
Guardrails AI
Deterministic structured output validation, JSON enforcement & safety filters
PyTorch
Deep learning tensors, GPU training, LoRA fine-tuning & Unsloth acceleration
HG Transformers
Open-source model inference, tokenization pipelines & vector embeddings
Hybrid RAG & Vector DBs
Dense + Sparse hybrid retrieval, Cohere reranking & semantic caching
Vapi & Voice AI
Real-time bidirectional ultra-low latency voice agents & STT/TTS
Temporal.io
Durable distributed execution, retry state machines & long-running jobs
GitHub Activity & Stats.
Real-time live GitHub metrics, active open-source contributions, and daily commits across intelligent AI frameworks.
In the past 12 months
Current daily streak
PRs merged & submitted
TypeScript & Python
Contribution Graph(Syncing live...)
1,804 contributions in the last year
Pinned Repositories
Open Source Projectsclario-career_platform
Personalized career acceleration platform featuring tailored roadmaps and AI-proctored mock interviews.
Wekraft-collaboration-platform/wekraft-saas
Wekraft - AI-First Project Management Platform. An intelligent, modern collaboration platform built with TypeScript & MCP.
Complete_Agentic_system
Autonomous multi-agent system architecture featuring cyclical LangGraph state machines, persistent memory, and MCP integrations.
customer_agent_punjabi
Enterprise bilingual LangGraph sales agent architected with semantic caching, hybrid RAG, guardrails, and persistent memory.