Hello, I'm

Viraj Vilas Annapurna Sawant

Full Stack Developer | AI Engineer

Download Resume

About Me

Passionate Full Stack Developer and AI Engineer with a strong focus on building scalable web applications and advanced AI models. Experienced in developing production-ready RAG chatbots, multi-agent systems, and deep learning models. Proven track record in hackathons and research programs.

Experience

Software Engineer Intern

CoderTrails

Sep 2024 – Apr 2025

Engineered a production-ready, AI-powered RAG chatbot using FastAPI, LangChain, and AWS for accurate retrieval and summarization from large documents. Boosted PDF query efficiency by 70%, slashing document analysis time.

Research Intern

Dr. Balasaheb Sawant Konkan Krishi Vidyapeeth

Jan 2023 – May 2023

Developed and optimized deep learning models (VGG-16, MobileNet, DenseNet) for crop disease detection. Improved detection accuracy by 15% and achieved over 90% real-world accuracy.

Research Fellow

Algoverse AI Research Program

Fall 2025 - Present

Selected for a competitive, mentor-led research program focused on advanced AI topics.

Backend Engineer

Freelance (Confidential Client)

Freelance

Designed and implemented a multi-agent AI system using LangChain and LangGraph. Optimized application performance by redesigning agent architecture to a parallel execution model.

Featured Projects

Codetrace-ai

Built a privacy-first autonomous code intelligence agent combining Vector DB (ChromaDB) and structural Graph DB (SQLite + NetworkX) for deep codebase understanding. Engineered hybrid semantic + call-graph analysis (Tree-sitter) enabling blast-radius impact analysis and cross-file dependency tracing across 6+ languages. Developed CLI + MCP server integration for IDEs (Cursor/Claude), supporting offline execution with local LLMs (Ollama) and smart delta re-indexing.

PythonTree-sitterChromaDBSQLiteNetworkXMCPOllama

Research Copilot — AI Research Assistant

Built a production multi-agent research assistant (LangGraph) with 3 specialized agents: Paper Analysis, Web Research, and SOTA Tracker to automate deep technical literature reviews. Designed a Reasoning RAG pipeline combining section-aware and hierarchical summarization with ReACT logic, enabling iterative reasoning traces to surface technical gaps and novel contributions in papers. Implemented an Advisor-Specialist agent validation layer that cross-checks technical accuracy across hierarchical document levels before report generation, significantly reducing hallucination risk.

LangGraphRAGReACTMulti-AgentNLP

PEAR — Personalized Evaluation of Allergens and Recipes

GitHub Hackathon Project. Rapidly prototyped an AI-driven MVP using LLaMA 3.2 to evaluate potential allergens in recipes, enhancing dietary safety.

LLaMA 3.2AIHackathon

SHL Assessment Gen-AI

Hugging Face Assignment. Automated recruiter workflows by building a Gen-AI tool to map job descriptions to assessment tools, reducing manual effort by 80%.

ReactFastAPILLaMA-3

AKD Daginawala

Engineered a full-stack jewelry discovery application using Flutter and Supabase for seamless cross-platform performance. (Under Development)

FlutterSupabaseMobile

Fake Emotion Detection System

Patent published. A system for detection of fake and genuine emotions from facial images. Also published a journal paper on CNN-ANN fusion approach.

ResearchComputer VisionPatent

Skills & Expertise

languages

PythonC++JavaScriptSQL

frameworks

ReactNext.jsFastAPIFlutterTailwind CSS

ai ml

TensorFlowPyTorchLangChainLangGraphOpenCV

tools

GitDockerAWSGCPJenkinsMongoDB

Get In Touch

Let's work together

I'm currently looking for new opportunities. Whether you have a question or just want to say hi, I'll try my best to get back to you!

+91-99678 92450
Mumbai, India