Skip to content
View Chai4156's full-sized avatar
🙂
Working from home
🙂
Working from home

Block or report Chai4156

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Chai4156/README.md

Hey, I'm Chai 👋

CS Undergrad · ML Engineer ·

LinkedIn · GitHub · Email


About Me

Final-year Computer Science student at Matrusri Engineering College, Hyderabad (CGPA: 7.9), actively seeking Data Science / ML Engineering roles from June 2026.

I build end-to-end ML pipelines — from raw data and model training to REST API deployment and frontend integration. My work spans NLP, computer vision, and recommendation systems. I care about understanding why a model works, not just that it does.

"RoBERTa scored 99.9% — and that's exactly why I didn't trust it."


🛠️ Tech Stack

Languages Python JavaScript SQL HTML CSS

ML / Data Scikit-learn Pandas NumPy

Frameworks Flask FastAPI React

Databases MongoDB PostgreSQL

Tools Git GitHub VS Code Jupyter Notebook

Core CS DSA OOP OS DBMS Computer Networks


🔬 Projects

NLP · RoBERTa · Naive Bayes · Flask · React · 2026

Built an ensemble combining Multinomial Naive Bayes + TF-IDF (84.8% accuracy) with a fine-tuned RoBERTa (99.9% in-distribution), achieving a 97.1% F1 score on the WELFake dataset (72,134 articles).

Key finding: RoBERTa's near-perfect score reflects stylistic overfitting, not semantic understanding — which is precisely what motivated the hybrid design. Deployed with a Flask REST backend and React frontend serving real-time predictions with per-prediction confidence scores.


Collaborative Filtering · Content-Based · MyAnimeList · 2025

Designed a hybrid recommender over 17,000+ MyAnimeList titles using KNN-based collaborative filtering combined with cosine similarity content-based filtering. Features include:

  • Genre-first onboarding to solve the cold-start problem
  • A dislike feedback loop that progressively filters similar candidates from future results

📜 Certifications

  • Google AI Essentials — 2025
  • English for IT — Cisco Networking Academy, 2025
  • Python for Everybody — University of Michigan, 2023

📊 LeetCode

Leetcode Stats


🌱 Currently Focusing On

  • Data Science fundamentals & Python depth
  • Consistency and deliberate practice

Available for full-time roles from June 2026 · Open to relocation

Popular repositories Loading

  1. Chai4156 Chai4156 Public

    About me

  2. Content-Recommender-System Content-Recommender-System Public

  3. hybridfakenews hybridfakenews Public

    a comprehensive fake news application with advanced scraping and ensemble approach.

    Python