Open to AI/ML & Software Engineering Roles

Engineering Intelligence.
Building What's Next.

I build AI-powered products and software that turn complex problems into practical solutions. Specializing in computer vision, generative AI & RAG pipelines, agentic workflows, and full-stack system architecture.

Mitul Jain
AI/ML Engineer · Software Developer
Pune, India
AI/ML GenAI Computer Vision RAG Systems Agentic AI Full-Stack ⚡ Core CS & DSA 🧠 Deep Learning
View Major Projects Contact Me
Think Deeply · Build Relentlessly
18.5204° N, 73.8567° E
Mitul Jain — AI/ML Engineer
MIT WPU '28

Mitul Jain

AI/ML Engineer

B.Tech CSE (AI & Data Science) · MIT-WPU Pune · Published Author in IJCDS

GitHub ↗ LinkedIn ↗ TakeUForward ↗
Academic Standing
9.11
CGPA out of 10.0 · B.Tech CSE (AI & Data Science)
Industrial Vision
97.76%
Inspection accuracy achieved at Tata Technologies
Vector Store Scale
700K+
Enterprise vectors indexed in Qdrant hybrid retrieval
RAG Retrieval
91%
Tested retrieval accuracy across 7,000+ real documents
DSA & Core CS
250+
Solved Problems on TakeUForward · Striver's Sheet
Research Publication
1 Paper
Peer-Reviewed in IJCDS · Sub-100ms Gesture Translation
01 · About

Built with engineering discipline, from research to production.

I'm an AI/ML engineer and software developer focused on building intelligent, useful, and scalable applications. I work across machine learning, generative AI, computer vision, and full-stack development — from experimentation to implementation.

My Engineering Methodology
SYSTEMATIC END-TO-END FLOW
STAGE 01
INPUT & SCOPE
Problem
Rigorous constraint mapping, failure mode anticipation, and domain data profiling.
STAGE 02
GRAPH ARCH
Design
Model & agent graph architecture, latency budget calculation, and API schema design.
STAGE 03
CODE & TRAIN
Build
Clean implementation in Python/C++, modular service decoupling, and reproducible pipelines.
STAGE 04
METRICS EVAL
Evaluate
Empirical benchmarking against mAP, IoU, retrieval precision, and edge-case testing.
STAGE 05
PROD INFERENCE
Deploy
Dockerized packaging, on-prem air-gapped readiness, and continuous monitoring.
02 · Real-World Engineering

Internship & Industry Collaboration.

Hands-on work in production environments — from a completed industrial computer vision collaboration with Tata Technologies to an ongoing AI internship engineering real-world models.

● Industry Collaboration

Agentic AI Intern · Visual Inspection Agent

Tata Technologies
January 2026 – April 2026 · Pune, India · Completed
  • Developed an on-premise industrial YOLOv8 Visual Inspection Agent using Python and OpenCV to detect and classify manufacturing defects in real time.
  • Engineered an agentic decision layer that automates inspection, re-inspection, and human escalation decisions, removing manual bottlenecks.
  • Achieved 97.76% inspection accuracy under harsh factory illumination, sensor noise, and class imbalance.
  • Architected as a secure, fully air-gapped software pipeline with zero cloud data egress.
Detector
YOLOv8 · Real-Time
Verified Accuracy
97.76% On-Floor
View on GitHub ↗ Watch Video Demo ▶
● Ongoing Internship

Artificial Intelligence Intern

WeIntern Pvt. Ltd.
September 2026 – Present · Pune, India / Remote
  • Developing production-ready AI solutions across Machine Learning, Deep Learning, Computer Vision, Generative AI, RAG, and Agentic AI.
  • Conducting rigorous empirical experimentation, algorithmic research, and pipeline optimization using modern Python AI frameworks.
  • Implementing modular retrieval architectures and autonomous agent tools to resolve real-world problem statements.
  • Collaborating with cross-functional engineering teams on model optimization, code reviews, and milestone deliverables.
Focus
AI / ML / GenAI
Status
Active · Remote
LinkedIn Profile ↗
03 · Peer-Reviewed Research

Published work.

Peer-reviewed research focused on making real-world computer vision systems accessible, robust, and low-latency.

IJCDS · August 2026 · Peer-Reviewed International Journal

Real-Time Sign Language Recognition Using CNN and MediaPipe with Enhanced Background Adaptation and Contextual Translation

International Journal of Computing and Digital Systems (IJCDS) · Vol. 19, Iss. 1 · Author: Mitul Jain

  • Architected a real-time gesture translation system combining Convolutional Neural Networks with MediaPipe landmark extraction achieving 97–99% recognition accuracy.
  • Engineered an adaptive background suppression & illumination normalization mechanism that prevents accuracy degradation in cluttered real-world environments with sub-100ms latency.
  • Implemented contextual translation converting isolated signs into grammatically coherent output for fluid communication.
97–99% Accuracy < 100ms Latency MediaPipe CNNs Accessibility AI IJCDS Vol. 19
Read Full Paper (IJCDS) ResearchGate ORCID Profile
04 · Flagship Systems

Major Engineering Projects.

Four flagship builds showcasing multi-agent orchestration, explainable inference, private offline RAG, and deep computer vision.

PROJECT 01 · AIR-GAPPED ENTERPRISE AI OPERATING SYSTEM 2026

SecureGPT — Air-Gapped Enterprise AI OS (No-Internet · LLM-Agnostic)

A sovereign, zero-internet enterprise intelligence platform engineered to autonomously orchestrate company workflows — completely air-gapped, on-premise, and LLM-agnostic.

  • Zero-Internet Data Boundary: 100% on-premise execution with Ollama; zero queries, tokens, or enterprise files ever leave secure local infrastructure.
  • Company-Wide Task Orchestration: LangGraph multi-agent graph coordinates specialized agents for document analysis, sandboxed code, and vision tasks.
  • Hybrid Enterprise Search: 91% retrieval accuracy using dense vector search and BM25 across 700K+ vectors indexing 7,000+ real documents with strict RBAC.
Deployment
Air-Gapped · No-Net
LLM Engine
LLM-Agnostic (Ollama)
Retrieval Acc.
91% (7K Docs)
Air-Gapped AI LangGraph LLM-Agnostic Qdrant FastAPI Docker
View on GitHub Watch Video Demo
PROJECT 02 · EXPLAINABLE AI 2026 · IIIT Pune

XplainAI — Explainable Multi-Agent Decision System

An 11-stage agent graph engineered for the IIIT Pune Hackathon (PS01) providing an interactive reasoning tree that allows humans to audit every intermediate decision step.

  • 11-Stage Reasoning Graph: Sequences intent classification, missing context resolution, task planning, live financial/news ingest, and candidate scoring.
  • Interactive Decision Visualizer: React frontend visualizes decision paths, clearly explaining why candidate answers were selected or pruned.
  • Auditable LLM Governance: Replaces black-box opacity with verifiable provenance, ensuring all outputs trace back to grounded data.
Pipeline
11-Stage Graph
Frontend
React.js
Auditability
Reasoning Tree
Explainable AI Multi-Agent React.js FastAPI Python LangGraph
View on GitHub Watch Video Demo
PROJECT 03 · OFFLINE RAG LEARNING ASSISTANT 2025

PyMentor — RAG-Based AI Learning Assistant

An offline, privacy-first tutoring application that transforms lecture videos into conversational, source-grounded study material without cloud subscriptions or external APIs.

  • Zero-Cloud Whisper Audio Pipeline: Transcribes lecture audio via Whisper and computes semantic embeddings with BGE-M3 into a local vector store.
  • Local On-Device Inference: Grounded cosine-similarity retrieval queries an on-premise LLaMA 3.2 model via Ollama with ~1.3s response turnaround.
  • Zero Data Egress Guarantee: Enforces air-gapped student privacy with 0 telemetry or queries transmitted outside the local machine environment.
Transcription
Whisper
Embeddings
BGE-M3
Latency
~1.3s Offline
Whisper BGE-M3 LLaMA 3.2 Ollama NumPy Index Python
View Repository
PROJECT 04 · COMPUTER VISION Ongoing Development

Cattle Weight Prediction via Vision Models

A deep-learning-based non-invasive livestock weight estimation system using multi-angle visual imagery to eliminate physical scale stress and operational livestock bottlenecks.

  • Contour & Landmark Extraction: Applies computer vision segmentation to extract precise anatomical contours and volumetric proxies from multi-angle RGB feeds.
  • Deep CNN Regression Architecture: Trains deep regression networks to map dimensional anatomical features directly to biomass weight with calibrated error bounds.
  • Edge Agritech Deployment: Engineered for contact-free monitoring in precision agriculture and livestock farm management workflows.
Status
In Dev 🚀
Task
Morphometric CV
Modality
Multi-angle RGB
Computer Vision Deep Learning CNN Regressor OpenCV Python
Explore Codebase
Additional GitHub Codebases & Builds
View All on GitHub (@mituljn14) ↗
Python 19 Business Intents

CompanyMate AI Assistant

Intelligent enterprise customer & business support assistant powered by LangGraph, FastAPI, Streamlit, and Google Gemini with deterministic escalation.

LangGraph FastAPI Streamlit Gemini Python
GitHub ↗
Python Dynamic Routing

LangGraph AI Content Generator

Modular AI content generation engine powered by LangGraph, FastAPI, Streamlit, and Google GenAI (Gemini) with iterative prompt refinement.

LangGraph Gemini API FastAPI Streamlit GenAI
GitHub ↗
Python Privacy-First ATS

TalentLens AI Resume Screening

Local, privacy-first AI resume screening and candidate ranking system powered by LangGraph, BGE-M3 Dense Embeddings, Ollama, and FastAPI.

LangGraph BGE-M3 Ollama FastAPI Streamlit
GitHub ↗
Python 97.76% Acc

Visual Inspection Agent

Industrial YOLOv8 model for on-floor defect classification (Tata Technologies) and automated agentic decision routing.

YOLOv8 OpenCV Agentic AI
GitHub ↗ Demo ▶
HTML / SQL Full-Stack

College Registration Portal

Transactional enrollment system built with Flask, MySQL, and PL/SQL for ACID-compliant student records.

Flask MySQL PL/SQL
GitHub ↗
Jupyter 1901-2015 Data

Crop Yield & Rainfall Predictor

Machine learning regression forecasting crop yields and optimal seed allocation based on century climate metrics.

Scikit-learn Flask Pandas
GitHub ↗
JavaScript React App

NewsMonkey Aggregator

Component-based React news application aggregating categorized live global news feeds with instant filtering.

React.js REST APIs Bootstrap
GitHub ↗
Python Open-Source

Solace Agent Mesh (Contrib)

Contribution to an event-driven framework for orchestrating distributed multi-agent AI systems against real data feeds.

Multi-Agent Event-Driven Python
GitHub ↗
Python MLOps

MLflow Experiment Tracking

Structured experimentation pipelines tracking model hyperparameters, loss curves, and artifact versioning.

MLflow Model Registry Python
GitHub ↗
05 · Technical Arsenal

Verified technical skillset.

Organized by engineering discipline. Grounded strictly in production projects, research publications, and demonstrated codebases.

⚡
Core Computer Science & DSA Problem Solving
Engineering fundamentals · Striver's A2Z DSA Sheet · 250+ Solved Problems
TakeUForward Profile (250+ Solved)
Data Structures & Algorithms (DSA) Operating Systems (OS) Database Management Systems (DBMS) Object-Oriented Programming (OOP) Computer Networks (CN) System Design Principles Striver's A2Z DSA Sheet Dynamic Programming Graph Algorithms & Trees Recursion & Backtracking Time & Space Complexity (Big-O) SQL Query Optimization ACID Transactions & Indexing Process Synchronization & Deadlocks Virtual Memory & Paging TCP/IP & HTTP Protocols
🧠
AI & Machine Learning
Supervised Learning Regression & Classification Clustering Feature Engineering Model Optimization Scikit-learn MLflow Cross-Validation
👁️
Deep Learning & CV
CNN Architectures YOLOv8 OpenCV MediaPipe Object Detection Image Segmentation TensorFlow / Keras mAP · IoU · Recall
⚡
Generative AI & RAG
RAG Pipelines Hybrid Dense Retrieval BGE-M3 Embeddings Qdrant Vector Store Ollama Local LLMs LLaMA 3.2 Whisper Audio Prompt Engineering
🤖
Agentic AI & Workflows
LangGraph Multi-Agent Orchestration Agent Routing & Tools Autonomous Graph Flows State Management Decision Boundary Logic Multi-Level Verification
🌐
Full-Stack & MERN Frameworks
MERN Stack React.js Node.js Express.js MongoDB RESTful APIs Component Systems State Management Vite / Modern Tooling
⚙️
Backend & Infrastructure
FastAPI Flask Node.js Express.js REST APIs Microservices Docker Containers MySQL MongoDB PL/SQL Redis Caching On-Prem Air-Gapped Setup Git & GitHub CI/CD Linux / Bash
💻
Programming & Core Data
Python C++ SQL JavaScript HTML5 / CSS3 NumPy Pandas Joblib FFmpeg
06 · Academics & Credentials

Formal education & verified certifications.

2024 – 2028 (Expected)
B.Tech CSE — AI & Data Science
MIT World Peace University, Pune, India
CGPA: 9.11 / 10.0
2023
Senior Secondary (Class XII)
Ascent International Sr. Sec. School, Udaipur
Score: 81%
2021
Secondary Education (Class X)
Lav Kush Sr. Sec. School, Salumber
Score: 98%

Verified Honors & Certifications

Smart India Hackathon (SIH)
Top 100 Teams University-Wide · MIT-WPU Round
99-Hr Data Science & Deep Learning
Krish Naik · KRISHAI Technologies (Udemy)
Data Analytics Job Simulation
Deloitte Australia · Forage
AI Professional Certificate
Unified Mentor
McKinsey.org Forward Program
McKinsey.org Leadership & Problem Solving
GitHub Fundamentals
Microsoft
Python for Data Science
Core Algorithms & Numerical Computing
Full-Stack Web Development
Modern Web Architecture

Let's build something scalable.

Available for AI/ML engineering roles, computer vision development, RAG systems, and full-stack software collaborations. Reach out directly — I respond promptly.

+91 86198 77759 GitHub Profile LinkedIn Profile TakeUForward (250+ DSA)
Email copied to clipboard!