I’m a B.Tech Chemical Engineering student at IIT Jodhpur with a Minor in Artificial Intelligence, interested in building software that combines strong engineering fundamentals with modern AI.

My work spans several layers of the stack:

• AI systems — LLM orchestration, RAG, agents, memory, tool use, vision pipelines
• Backend & distributed applications — APIs, realtime systems, databases, caching, authentication
• Systems programming — C++, Rust, Linux, networking, native applications
• Applied ML & scientific computing — computer vision, deep learning, process modeling and reinforcement learning

I enjoy projects where the difficult part is not just calling a model, but designing the system around it: data flow, state, memory, reliability, observability, and the interface between software components.

Currently, I’m focused on AI Engineering and SDE roles, while continuing to explore systems programming and applied research.

Research Intern · Global Optimization and Knowledge Unearthing Laboratory

Worked on a scientific-document intelligence workflow for extracting structured information from research papers.

The pipeline combines:

PDF / Document Parsing → Layout Detection → OCR → Relationship Extraction → Figure Understanding

Technologies explored include YOLO, Tesseract OCR, LaTeX OCR, vision-language models, Python, React, Vite, and Electron.

A major focus was extracting quantitative information from scientific figures, particularly temperature vs. conversion/selectivity curves, while preserving the experimental conditions, catalysts, promoters, captions, and relationships needed for downstream retrieval and analysis.

Rust · Python · LangGraph · Neo4j · Qdrant · MCP

A systems-oriented AI assistant designed around the idea that an assistant should be more than a chat interface.

Architecture highlights

• Rust terminal client for the interactive shell
• Python runtime for orchestration and AI workflows
• Async IPC over Unix sockets
• LangGraph-based agent coordination
• Neo4j for structured / relational memory
• Qdrant for semantic retrieval
• MCP-based tool integration
• Modular services for memory, reasoning, execution, and observability

The goal is to build an assistant whose state, memory, tools, and reasoning are explicit system components rather than hidden behind a single model call.

Python · YOLO · OCR · Vision Models · React · Vite · Electron

A desktop research tool built during my IIT Hyderabad internship to turn unstructured scientific PDFs into structured, machine-readable research data.

The system focuses on scientific figures and their surrounding context, including:

• figure / chart detection
• layout-region extraction
• OCR and mathematical text extraction
• figure-caption relationships
• vision-model interpretation
• extraction of plotted data and experimental conditions
• structured outputs suitable for search and RAG pipelines

C++ · CMake · SQL · Storage Engines

A lightweight SQL database engine built from scratch as a systems project.

Implemented components include:

• SQL parsing and execution
• record management
• indexing
• table catalogs
• disk-based storage
• memory management
• debugging infrastructure

This project is part of my interest in understanding software below the application layer — how databases, storage, and execution engines actually work.

C++17 · Qt 6 · CMake · QGraphicsView

A desktop chess application with a native GUI and a complete rules engine.

Supports:

• legal move validation
• check / checkmate
• stalemate
• castling
• en passant
• pawn promotion
• move highlighting
• scalable board rendering

A useful combination of object-oriented design, graphics programming, game-state modeling, and rule validation.

Python · PyTorch · Reinforcement Learning · Process Control

An applied RL project exploring automatic PID tuning for nonlinear process systems.

The project reproduces and extends a TD3-based controller-tuning approach with techniques such as:

• modified reward design using control-rate penalties
• prioritized experience replay
• curriculum learning
• evaluation across multiple plant models
• comparison against classical control behavior

It sits directly at the intersection of my Chemical Engineering and AI background: dynamic systems + control theory + reinforcement learning.

Django · PostgreSQL · React · Electron · WireGuard · Paramiko

A full-stack VPN management system with a desktop client and remote server management.

The project brings together:

WireGuard ↔ Django REST API ↔ PostgreSQL ↔ Electron/React Client ↔ SSH Infrastructure

with features for peer management, VPN configuration, connection statistics, and remote server administration.

Python · LangChain · Local LLMs · Ollama · RAG

An early exploration into autonomous research workflows: taking a high-level objective, decomposing it into tasks, searching for information, and synthesizing results through an agentic workflow.

ARAA became one of the foundations for my later interest in agent orchestration, tool use, memory, and research automation.

LangChain · LangGraph · RAG · Computer Vision · OCR · Reinforcement Learning · LLM Agents · Ollama

Neo4j · Qdrant · REST APIs · WebSockets · Unix Sockets · CMake · WireGuard

                ┌─────────────────────────┐
                │      AI / Reasoning     │
                │  LLMs · Agents · RAG    │
                └────────────┬────────────┘
                             │
                ┌────────────▼────────────┐
                │      Application        │
                │ APIs · Realtime · UX    │
                └────────────┬────────────┘
                             │
                ┌────────────▼────────────┐
                │        Systems          │
                │ C++ · Rust · Linux      │
                └────────────┬────────────┘
                             │
                ┌────────────▼────────────┐
                │   Scientific Domains    │
                │ Process · ML · Research │
                └─────────────────────────┘

The common thread across these projects is turning ideas into working systems — not just prototypes, but architectures that can be reasoned about, extended, and operated.

A space for building **AI systems, developer tools, infrastructure, and experimental software**.

Automatically regenerated by the repository's pacman.yml workflow.

I also maintain smaller projects and experiments across:

DSA · Computer Vision · NLP · Web Applications · Qt · Networking · Control Systems · Deep Learning · Developer Tooling

Some of these are deliberately small — they are experiments for learning a technology, testing an idea, or understanding an implementation from first principles.

I’m interested in AI engineering, software engineering, systems, developer infrastructure, and applied research.

LinkedIn · LeetCode · GitHub · Limbo Corps