Hello!

I'm Somesh Shukla



YouTube My works

Resume

Resume

Software Engineer with expertise in Generative AI, LLMs, and the MERN stack. Skilled in designing intelligent applications, RAG systems, and conversational AI solutions

Experience


Oct 2025 – Present

Generative AI Engineer

Finkraft, Bengaluru, Karnataka

  • Built a high-precision Retrieval-Augmented Generation (RAG) pipeline for a hotel invoice processing system, utilizing Multimodal LLMs to extract structured data from complex unstructured documents.
  • Implemented a RAG repair pipeline utilizing Vector Stores (FAISS) and MongoDB; researched and integrated Graph RAG concepts to model complex entity relationships, increasing data parsing accuracy from 30% to 99%.
  • Designed the orchestration layer for an Agentic AI system using Python and LangGraph, enabling the LLM to autonomously connect to external APIs and manage intelligent escalation paths for enterprise workflows.

Jul 2025 – Aug 2025

Generative AI Engineer Intern

Duco Consultancy, Gurugram, Haryana

  • Developed a scalable hospital management system featuring an AI Agent with RAG capabilities; optimized the retrieval architecture to ensure secure access to medical records while maintaining strict data privacy.
  • Implemented agentic tool-use to integrate LLMs with wearable device APIs, automating real-time health data tracking and appointment scheduling.

Apr 2024 – Sep 2024

Software Engineer Intern

Meta XR, Kolkata, West Bengal

  • Developed an AI-driven VR interview platform using Unreal Engine and C++, integrating interactive MetaHumans driven by LLM backends and engineered agentic behavior using the Convai SDK for dynamic user interactions.

Personal Experience

AI & Full Stack Development

My personal development experience includes various AI applications and full stack solutions that I have worked on throughout my college life.

  • Worked on LLM-based applications including chatbots, document analyzers, and intelligent agents.
  • Built AI-powered web applications with real-time data processing and machine learning integration.
  • Developed intelligent recommendation systems and automated content generation tools.



Education


2021-2025

Bachelor of Technology

Chandigarh Engineering College

Grade: 7

2019-2020

Higher Secondary School

Kendriya Vidyalaya

Grade: 80%

Download CV

Projects

Projects

Below are AI-powered applications and intelligent systems built using cutting-edge GenAI technologies, LLMs, and modern web frameworks.

Visual RAG for Layout Matching

  • Converted documents to images, embedded them using Vertex AI multimodal embeddings, and searched a FAISS index for the most visually similar previously-parsed document.
  • Injected scrubbed JSON schema (values replaced with placeholders) as a structural map; combined with spatial-relational OCR using full bounding box coordinates [x, y, w, h] to reason about field positions, eliminating layout misalignment and hallucinations.

Self-Hosted Multimodal Chatbot & RAG Repair Agents

  • Created RAG repair agents for high-volume document parsing pipelines using embedding models, vector databases (FAISS), and Graph RAG; also architected autonomous email workflow agents and NotebookLM-style AI research assistant for private document analysis and PRD generation.
  • Built internal chatbot with self-hosted Qwen 2.5 multimodal model; swapped Gemini API with Chandra OCR inference (hosted on VM) and fine-tuned DeepSeek OCR 2 on company data.
  • Added agentic capabilities including ag-grid frontend manipulation, one-click report downloads, and full knowledge-base integration.

Inquestor : AI Agentic Research

Inquestor is an AI Agent built for research, featuring an interactive and user-friendly web interface built with Next.js and React that allows users to input queries, view structured results with sources, and download the generated research as a PDF document.

AI VR Interview Platform Using Conversational AI

An AI-powered VR Interview Platform leveraging conversational AI to simulate realistic interview scenarios. This platform helps users practice and refine their communication skills through interactive dialogues with AI-driven virtual interviewers, providing instant feedback and performance analysis.

A RAG app to ask queries about multiple PDFs simultaneously

An interactive PDF Chat Application built with Streamlit, allowing users to upload PDF documents and ask questions about their content. The app leverages Google's Generative AI (Gemini for chat, and embedding models for RAG) and LangChain to process PDFs, create FAISS vector stores, and provide context-aware answers. Features include multi-PDF management and persistent local vector storage.

More projects on Github

I love to build AI-powered applications with modern tech stacks & intelligent systems


GitHub

Skills

Python 90%
Generative AI 85%
LangChain & RAG 80%
Machine Learning 75%
Javascript 80%
MERN Stack 75%
AWS 100%
Vector Databases 70%

About

About Me

With expertise in Generative AI and modern web technologies, accompanied by a bachelor's degree in engineering, I bring proficiency in AI/ML, conversational AI systems, and full stack development. Skilled in crafting intelligent applications, RAG systems, and implementing scalable AI solutions. Demonstrated success in managing AI-driven projects and collaborating effectively with teams to deliver innovative solutions.

  • Profile: Software Engineer
  • Domain: GenAI, RAG Systems, MERN Stack & AI Chatbots
  • Education: Bachelor of Technology
  • Language: Python, Javascript, C++
  • AI Tools: LangChain, OpenAI APIs, Streamlit & Vector DBs
  • Other Skills: React.js, Node.js, MongoDB, Express.js, SQL, Git & JIRA
  • Interest: AI Research, Machine Learning, Traveling

LinkedIn

Contact

Contact Me

Below are the details to reach out to me!

Address

Bengaluru, India

Contact Number

+ 918423069420

Email Address

somesh13shukla@gmail.com

Download Resume

Drive link



  • Find me on

Somesh Shukla's Portfolio AI Assistant
Hello! I'm Somesh's AI assistant. How can I help you today?
💡 First response may take ~30s as the server wakes up.