AI Engineer · Pune, India

OmkarShinde

I love building things with AI — experimenting with language models, designing intelligent systems, and turning wild ideas into working software.

Currently obsessed with: |

See What I've Built → Say Hello Download Resume ↓
Scroll

Who I Am

Builder.
Tinkerer.
AI Nerd.

1+
Years in the industry
15+
Technologies used
∞
Tabs open at any time
☕
Runs on coffee & curiosity

I'm Omkar — a Software Engineer at CGI in Pune who got completely hooked on AI and never looked back. I studied Computer Science with a focus on AI at MIT Manipal, and that rabbit hole just kept getting deeper.

What actually gets me excited? Building things that think. I love taking a blank canvas and figuring out how to make a machine do something intelligent — whether that's parsing documents, understanding code, automating workflows, or generating insights nobody asked it to generate but everyone needed.

I'm the kind of person who builds a personal project at midnight just to see if an idea works. My side projects have involved recommendation engines, resume parsers, OCR pipelines — all because I was curious about whether it could be done.

Right now I'm levelling up hard — grinding DSA, going deep on AI system design, and targeting an AI Engineer role where I can do this full-time.

Work

Where I've
Been Building

Software Engineer · CGI
Sep 2025 – Present
📍 Pune, Maharashtra
  • Building LLM-powered automation systems that make real enterprise workflows faster, smarter, and less painful for developers and stakeholders alike.
  • Designing and shipping end-to-end AI pipelines — from data extraction and knowledge graph construction to intelligent summarisation and automated code analysis.
  • Working at the intersection of AI engineering and backend development — making sure the models don't just work in a notebook, but actually run reliably in production.
Data Science Intern · ELIXIR.AI
Mar 2024 – Jul 2024
📍 Pune, Maharashtra
  • Worked on computer vision and OCR systems for real-world document processing — messy handwritten data, tight accuracy requirements, actual production constraints.
  • Trained and deployed custom neural networks for compliance monitoring, hitting 95% precision on challenging real-world image datasets.
  • Got a solid lesson in the difference between "works in a notebook" and "works in production."

Personal Work

Things I've
Built

01 / Personal
AI Resume-to-Job Matching Platform
Built an end-to-end RAG pipeline — resume PDF upload → ML role classification → vector embedding with SentenceTransformers → ChromaDB vector store → live job fetching via SerpAPI → semantic similarity scoring → ranked results with match percentages. Integrated Claude for resume analysis and email delivery of results.
PythonFlaskSentenceTransformersChromaDBSerpAPIspaCy
Accuracy 85%+Categories 9Data prep ↓90%
02 / Personal
Video Recommendation System
Deep learning recommendation engine with embedding layers, ReLU activations, and dropout regularisation. Integrated external social platform APIs for user viewing and rating signals, with cold-start handling via mood-based fallback recommendations.
FastAPITensorFlowScikit-learnSQLiteSQLAlchemy
Response <500msCold-start handled
03 / Internship
OCR for Handwritten Medical Records
Handwritten doctor prescriptions are basically a different language. Built a pipeline to extract structured data from them — because healthcare data should be accessible, not locked in illegible scans.
PythonPyTesseractOpenCVComputer Vision
Accuracy ↑20%Fields 8+ extracted
04 / Work
Codebase Intelligence System
An AI assistant that lets developers ask questions about a large legacy C++ codebase in plain English and get accurate, context-aware answers. Built with a knowledge graph in Neo4j and semantic retrieval — so it understands both code structure and meaning.
PythonNeo4jKnowledge GraphsLLMSemantic Search
Codebase 10,000+ linesOnboarding ↓40%
05 / Work
AI Report Migration Pipeline
Enterprise reporting systems are a pain to migrate manually. Built an AI pipeline that reads legacy reports, understands their formula logic, and automatically rewrites them in the target format — across 200+ templates.
PythonLLMPrompt EngineeringJSONAutomation
Effort ↓75%Templates 200+
06 / Work
AI Pull Request Review Bot
A GitHub bot that actually reads your pull request, spots functions without test coverage, writes the missing unit tests for you, and posts a structured review — all before a human even looks at it.
PythonGitHub APILLMAutomation
Review time ↓60%Saves 2–3 hrs/sprint
07 / Work
Stakeholder Report Generator
Nobody enjoys writing status reports. Built an automation that crawls the full project hierarchy, aggregates everything across tickets and releases, and generates a clean stakeholder-ready report — in under 2 minutes instead of 2 hours.
PythonJira REST APILLMSummarisation
Report time 2hrs → 2minTickets 50+/cycle

Toolkit

What I Work With

Languages & Core
Python
Java
JavaScript
HTML/CSS
SQL
AI & Machine Learning
TensorFlow
KerasKeras
Scikit-learn
OpenCV
spaCy spaCy
LANGCHAIN LangChain
GraphRAG Graph RAG
RAGVectors RAG / Embeddings
Neo4jgraphNeo4j
Frameworks & Backend
FastAPI
Flask
React.js
Node.js
pandas📊 Pandas
NumPy NumPy
Databases & Tools
MongoDB
MySQL
SQLite
GitHub
Bitbucket
VS Code
JiraJira
Git

Contact

Let's Build
Something

Open to AI Engineer roles. If you're working on something interesting — an AI product, a bold idea, or just want to nerd out about language models — I'd genuinely love to talk.

Email Me LinkedIn GitHub Resume