Hellow World!

I'm Aditya Kulkarni

>  

I build intelligent systems and ship them as real products.

Currently obsessed with ML pipelines and Flutter apps - because what's the point of training a great model if users never see it? I turn messy datasets into predictions and predictions into experiences people can actually use.

Final-year CS student at SSGMCE. Actively seeking ML Engineer, Data Engineer, and Flutter Developer roles.

ABOUT ME

I'm a final-year Computer Science student at SSGMCE with a thing for building end-to-end solutions—from raw data to deployed products.

My toolkit: Python for ML pipelines, Flutter for shipping apps, and Linux for everything in between. I've gone from a 1.2% R² baseline to 90.4% accuracy through ensemble modeling, built autonomous job scrapers that rank opportunities by priority, and digitized manual railway operations into real-time management systems.

Currently interning with Central Railway Engineers' Association, modernizing their workflow from Google Forms to a full-stack web portal. Previously interned at Ministry of Railways, where I built locomotive maintenance scheduling features.

What I'm looking for:

ML Engineer, Data Engineer, or Flutter Developer roles where I can build things that actually ship and solve real problems.

STACK_INVENTORY

Languages

Python Java Dart SQL

ML & Data

Pandas NumPy Scikit-learn Playwright ETL XGBoost CatBoost

Tools

Git Linux Jupyter SQLite Firebase

PROJECTS _

LinkedIn Job Intelligence Agent

In Progress • Python Pipeline

Autonomous job hunting system that scrapes LinkedIn daily, scores opportunities by a custom Priority Algorithm (recency × competition ratio), and delivers ranked reports straight to your inbox.

Playwright-based authenticated scraping with anti-detection

SQLite deduplication prevents duplicate tracking

Cron-scheduled automation for daily intelligence reports

Next: Adding NLP-based job description analysis to surface skill-matched roles

#Python #Playwright #SQLite #ETL #Automation

Stacking Regressor - Used Car Price Predictor

Production ML Model

Transformed a messy dataset into a production-ready prediction model, skyrocketing accuracy from 1.2% → 90.4% R² through aggressive feature engineering and ensemble stacking.

Heavy preprocessing: cleaned text features, engineered 15+ new variables

Trained 10+ models, compared performance rigorously

Final architecture: Stacking Regressor (RF + XGBoost + CatBoost)

Model predicts prices within ±5% accuracy—ready for dealership tools or consumer apps

#Scikit-learn #XGBoost #CatBoost #Pandas #NumPy

Loco-Info Railway Manager

Ministry of Railways • Production

Digitized decades of paper-based locomotive tracking for Central Railway's Bhusawal shed. Now tracks 50+ locomotives in real-time with maintenance scheduling, staff allocation, and operational status.

Reduced data retrieval from 30 minutes5 seconds

Firebase real-time sync for 5+ concurrent users

Built entire Flutter frontend (10+ screens)

Used daily by Electric Loco Shed operations team

#Flutter #Firebase #Dart #Real-time

CREA Web Portal

Ongoing • Full-Stack

Modernizing Central Railway Engineers' Association (500+ members) from scattered Google Sheets → centralized web platform with member management, event registration, and admin dashboards.

Authentication system + role-based access control

Dynamic forms with automated email notifications

Replacing manual Google Forms workflow entirely

Status: 60% complete • Member registration & event modules live in staging

#React #Node.js #MongoDB #Express
Launching Jan 2026

More experiments on GitHub & Kaggle

EXPERIENCE_LOG

Web Developer Intern

ONGOING

Central Railway Engineers' Association

Sept 2025 - Present

Modernizing workflow for 500+ railway engineers from scattered Google Forms → centralized web platform. Building member management system, event registration portal, and admin dashboards with authentication and automated email workflows.

Architected full-stack solution (React + Node.js + MongoDB)

Implemented role-based access control for admin/member roles

60% complete • Launching Jan 2025

App Developer Intern

COMPLETED

Ministry of Railways, Central Railway

Electric Loco Shed, Bhusawal • Aug 25 - Sept 25

Developed Loco-Info, a railway shed management system that digitized manual locomotive tracking for 50+ units. Built the entire Flutter frontend with real-time Firebase sync, reducing data retrieval time from 30 minutes to under 5 seconds.

Designed 10+ screens for maintenance scheduling and staff allocation

Implemented role-based access control for admin/member roles

Now used daily by shed operations team

BEYOND_CODE

When I'm not training models or debugging Flutter widgets, you'll find me deep in the FOSS ecosystem, hunting for tools that do one thing exceptionally well. I believe the best software is the kind you can peek under the hood of, which is why I daily-drive Arch Linux + GNOME. Minimal setup, maximum control.

I'm also an RPG enthusiast who believes the best engineers know when to step away from the screen. Whether that's exploring open worlds or actually touching grass, balance keeps the code quality high and the burnout low.

> random_facts.txt

  • • Switched to Linux and never looked back, tried Ubuntu, Fedora, settled on Arch
  • • Prefer terminal over GUI for most tasks (yes, even for Git)
  • • Current Favorite FOSS apps: Raindrop.io for bookmarks, Wave Terminal Terminal of choice, AppFlowy for notes
  • • Coffee >> Tea (this is non-negotiable)

Want to steal my setup? Hit me up for the full list.

If you've read this far, we're probably going to get along. Let's build something cool together.