I am an Electronics & Communication Engineering (ECE) student specializing in Agentic AI development, web development, and workspace automation. I focus on building remote download managers, self-hosted services, and intelligent multi-agent workflows to automate digital workspaces.

• 🔭 Current Focus: Developing multi-agent AI systems and full-stack web applications
• 🤖 Agentic Automation: Orchestrating task flows by integrating LLMs/LangChain with workspace APIs and local environments
• ⚙️ Systems & DevOps: Designing self-hosted configurations, local server networks, and customized desktop environments
• 📸 Creative Side: Photography enthusiast
• ⚡ Fun Fact: Passionate about "ricing" Linux setups (Hyprland / KDE Plasma / Cinnamon)

Software & Web Stack

Systems & DevOps

• Agentic AI: LangChain, Multi-agent workflows, StateGraph Executors
• Web Frameworks: FastAPI (Python API backend), React (Web client dashboard)
• Infrastructure: Docker containment, self-hosted services orchestration

• Platforms & SBCs: ESP32, Raspberry Pi, Arduino
• Interfaces & Protocols: BLE (Bluetooth Low Energy), MQTT, UART/Serial, Wireshark packet analysis

[Flagship Project]

A secure, local-first personal assistant built on a modular Dynamic State-Graph Multi-Agent Architecture (LangGraph). It decomposes complex user requests into structured, sequential subtasks and delegates them to sandboxed worker agents.

Key Architecture & Tech:

• Orchestration Engine: Cyclic state machine utilizing LangGraph to coordinate workers, prevent context pollution/drift, and resolve execution deadlocks
• Sandboxed Worker Agents: Independent ReAct workers (System, Gmail, Productivity, GitHub, and Google Classroom) equipped with target-specific Python tools
• Nested Obsidian Team: A specialized sub-graph team featuring an Obsidian Manager and sub-workers (Note, Canvas, and Refactor agents) with dynamic folder nesting
• Zero-Trust Security: High-risk OS and shell commands isolated behind a password verification layer and standard input mutex
• Context & Performance: Semantic vector search via FAISS (all-MiniLM-L6-v2) for skill discovery, and ChromaDB for user preference learning

Python LangGraph LangChain Gemini FAISS

A distributed, real-time remote download manager that separates download intent (submitted via web dashboard) from physical execution (run by Python agents on target devices).

Key Architecture & Tech:

• Event Bridge Sync: Combines Firestore (on_snapshot zero-polling dispatching) with Realtime Database (low-latency, sub-second progress updates and pause/resume/stop actions)
• Pluggable Download Engines: Uses the strategy pattern to coordinate Aria2 (multithreaded JSON-RPC downloads), yt-dlp (dynamic video extraction format selector), and direct HTTP/HTTPS engines
• Smart Categorization: AI-powered classifier dynamically routes downloaded files into organized folders (Media, Apps, Archives) based on URL and filename properties
• DevOps & Service Design: Configured for background execution as daemon systemd user services

React Vite FastAPI Python Firebase RTDB + Firestore Aria2

Designed for flexible electronics — secured 🥉 3rd Place in the Flexible Electronics track at the MakeInSilicon National Symposium.

Embedded Systems Flexible Electronics

Triggering notifications and device controls directly from PC via protocol analysis and Wireshark BLE captures.

BLE Protocol Analysis Python

"Building systems that work while you sleep."