Building AI-native systems, agent orchestration platforms, and intelligent developer tooling.
Focused on stateful reasoning, multi-agent coordination, and human-AI collaboration.
B.Tech Computer Science @ Muthoot Institute of Technology and Science
AI-native systems and agent orchestration platforms. Architected for stateful reasoning and human-AI collaboration.
MCP-native developer tooling that sits between AI coding assistants and git history. Quality gates, convention learning, and intelligent commit automation.
Key Systems:β’ MCP Server integration for Claude/Cursor/Windsurf
β’ Multi-provider AI orchestration (5 LLM providers)
β’ Deterministic quality gates (security, code smells, coverage)
β’ Convention learning engine from commit history
β’ TypeScript CLI + Next.js dashboard architecture
Stack: TypeScript, MCP SDK, Next.js, Supabase, Ollama
A state-based intelligence system that makes AI reasoning persistent, structured, and resumable. Intelligence = State. Reasoning = State Transitions.
Core Innovation:β’ Hybrid Cognitive Operators (HCO): Neural + Symbolic + Causal
β’ Persistent cognitive state across sessions
β’ Zero-latency context resumption (10-100x token reduction)
β’ 42/42 tests passing, Groq/Gemini/Ollama providers
β’ Causal dependency graph for reasoning traceability
Stack: Python, LLM APIs, State Machines
Transforms high-level goals into executable workflows. Compiles intent into structured AI systems with specialized agent dispatch and role matching.
Architecture Highlights:β’ 3-tier model orchestration (Quality/Efficiency/Speed)
β’ Dynamic agent-role matching with skill registry
β’ 6-dimensional quality scoring system
β’ Instruction file generation (CLAUDE.md, .cursorrules, AGENTS.md)
β’ Reactive workflow execution with rollback support
Stack: Next.js 15, TypeScript, OpenRouter, Groq, Framer Motion
Dual-mode fingerprinting system for real-time content protection. 3rd place at NEXUS Hackathon 2026 for digital asset protection innovation.
Technical Achievements:β’ pHash + dHash video fingerprinting with Mel-spectrogram audio
β’ 95% detection accuracy, 20x parallel speedup
β’ Groq Llama 3.3 70B for automated DMCA generation
β’ <90 second enforcement window
β’ React + Flask + WebSocket real-time dashboard
Stack: Python, OpenCV, React, Groq, Redis
| Role | Organization | Timeline | Focus |
|---|---|---|---|
| Tech Subcommittee Member | MITS Media Club | Oct 2024 β Present | Technical direction and infrastructure |
| Member | MITS Motorsports | Oct 2024 β Present | Engineering and team operations |
| Apprentice Trainee | Soften Technologies | May 2024 β Aug 2024 | Backend development (Python & JavaScript) |
- AI-native application architecture
- Multi-agent orchestration systems
- MCP (Model Context Protocol) integration
- Stateful reasoning and cognitive runtimes
- Developer tooling for the AI era
- Correctness-first system design