Gen AI  ·  Backend Engineer  ·  Gurugram, India

Mayank
Joshi.

Final-year IT student at ADGITM, wrapping up in June 2026. My heart genuinely belongs to agentic systems — and also to FC Barcelona (non-negotiable). When I'm not shipping code, I'm in the gym, out on the football pitch, or adding questionable tracks to a playlist that's honestly too good to share. I build at the intersection of LLMs and real backend engineering, RAG pipelines that actually retrieve the right context, agents that get things done, and APIs that don't make future-me cringe.

Get in touch See my work
Gurugram, India B.Tech IT · ADGITM '26 7.4 GPA
Mayank Joshi
Mayank Joshi
Gen AI · Backend Engineer
scroll
Work Experience
Oct'25 — Present
ML + Backend Engineer (Full-time, previously AI Intern)
Kosmoz Consulting ↗
  • Knowledge Graph-Enabled RAG System: Contributed to integrating a Neo4j knowledge graph into a RAG system to improve structured retrieval and context grounding.
  • End-to-End Product Architecture - Klex: Built the overall software architecture spanning backend, frontend, and upcoming AI features for a market-ready product.
  • Demand Forecasting: Developing a machine-learning demand-forecasting workflow that analyses historical demand patterns to support data-driven planning.
Python Neo4j Knowledge Graphs RAG Pipelines Machine Learning Demand Forecasting Backend Architecture
Feb'26 — Aug'26
AI-Agentic Systems Consultant
BNG Infotech Pvt Ltd. ↗
  • Database Agent: Developed an agentic database assistant that interprets natural-language requests, retrieves relevant structured data, and generates context-aware outputs.
  • Agentic Invoice Extraction: Built an agentic invoice-processing system that extracts structured fields from invoices and coordinates downstream processing workflows.
  • RAG Workflow Pipeline: Developed a retrieval-augmented generation pipeline for a workflow project, enabling context-grounded document retrieval and response generation.
Agentic AI Database Agents Document AI Invoice Extraction RAG Pipelines Python LLMs
Oct'24 — Feb'26
AI Engineer + Backend dev (Full-time, previously AI Intern)
Deeplogic AI ↗
  • PDF Translator Tool: Developed a complete end-to-end solution for translating German PDF documents into English, handling both frontend and backend implementation, and integrating a custom-trained translation model.
  • OCR Automation System: Engineered an OCR pipeline to extract and overlay machine-readable text from PDFs while preserving the original layout. Automated the integration with Django backend and frontend viewers, enabling faster document processing and improved accessibility.
  • Comic Production Automation: Contributed to automating the comic book production pipeline by leveraging AI.
Python Django REST APIs OCR Automation Translation Models Generative AI
Projects
✦ Live Release v1.0
WorkPilot AI — Local Agentic Work Orchestration Platform

A privacy-first, on-device AI assistant and multi-agent workspace orchestrator. Combines document intelligence, live desktop audio capture, and autonomous workflow execution with human-in-the-loop governance.

📄 DocOps (Document Intelligence)
On-device document parsing via Docling (PDF, DOCX, PPTX), entity extraction, document diffing, and hybrid RAG search backed by ChromaDB vector store.
🎙 MeetOps (Meeting Operations)
Pre-recorded file ingestion & real-time system/mic audio capture via Tauri v2 companion (WASAPI/PipeWire) with Faster-Whisper ASR & Pyannote diarization.
⚡ ActionOps (Agentic Execution)
Multi-agent task, blocker & decision extraction with Model Context Protocol (MCP) integrations (Jira, Slack, GDocs, Email) & 1-click HITL governance.
FastAPI Control Plane Tauri v2 + Rust React 19 + Vite Docling Faster-Whisper Pyannote.audio ChromaDB MCP Integrations OpenAI SQLite
🚀 Companion App Installers (Tauri v2 + Rust) Privacy-First · On-Device
🐧 Linux (.AppImage, .deb, .rpm) 🍏 macOS (.dmg) 🪟 Windows (.exe / .msi) All Releases ↗ GitHub Repository ↗
PlugTalk — LLM Chatbot

A dynamic chatbot that retrieves and provides answers based on scraped website content using conversational AI. Integrated custom API key functionality and real-time document retrieval via a RAG pipeline with Chroma embeddings.

LangChainOpenAIChromaStreamlitRAG
Live App: plugtalk.streamlit.app ↗
Football Q&A

Web application that answers queries related to football matches in real-time using natural language processing and LangChain.

LangChainOpenAINLP
Digital Cricket Community

Full-stack platform for cricket fans — posts, search, user profiles, and sponsorship features.

DjangoMySQLBootstrap
Legal Aid App & Chatbot

Free chatbot for basic legal assistance on abuse-related issues and informative blogs. 6th place out of 25 teams at Google DSC Hackathon.

PythonNLPDjango
Cloudburst Prediction

ML model taking location input to accurately predict cloudburst events with up to 65% accuracy. Built for Smart India Hackathon.

PythonMachine LearningSIH
Skills & Stack
Agentic AI & Machine Learning
Agentic AI Systems Generative AI Machine Learning LLM Applications RAG Pipelines Knowledge Graphs (Neo4j) MCP Tool Orchestration Database Agents Document AI & OCR Demand Forecasting LangChain OpenAI API Prompt Engineering Vector Search ChromaDB AI Workflow Orchestration Docling Faster-Whisper Pyannote Diarization
Backend & Architecture
Python FastAPI Django Flask REST API Design & Integration Backend Service Development Tauri v2 (Rust) Model Context Protocol (MCP) End-to-End Product Architecture
Databases & Data
SQL Neo4j Graph DB ChromaDB Vector Store MySQL SQLite Data Processing Query Optimization
Tools & Developer Stack
Git GitHub Linux Postman JavaScript HTML / CSS Streamlit
Smart India Hackathon 2024 — Cleared for National Finals
Smart India Hackathon 2023 — Top 35 teams in college round out of 100+ teams
Launchpad Ideathon (Jan '23) — Top 10 out of 100+ teams
Google DSC Hackathon (Dec '23) — 6th position out of 25 teams
Contact

Let's build something
together.

Looking for full-time roles and internships in AI engineering & backend. If you're building something interesting, I'd love to hear about it.

mayank.jo@outlook.com
Also on   GitHub  ·  LinkedIn  ·  Twitter  ·  Medium