Skip to content
View Debasish65368's full-sized avatar

Highlights

  • Pro

Block or report Debasish65368

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.

Content in all repositories owned by your account will be closed.
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
Debasish65368/README.md

Hi, I'm Debasish Kumar 👋

B.Tech CSE | Data Analyst • AI/ML Engineer • Backend Developer

Typing SVG


About Me

I'm a final-year Computer Science student who works with data end-to-end — from writing SQL and building ETL pipelines, to training ML models, to shipping the FastAPI/React applications that put those models in front of users.

What I do

  • Build full-stack, production-shaped systems — not just notebooks: deployed APIs, real dashboards, working demos
  • Work across the stack: SQL/ETL → model training & evaluation → backend API → frontend/dashboard
  • Focus on explainability and correctness — e.g. auditing an AI-generated SQL layer for real security bypasses, not just accuracy metrics

Currently

  • 🔭 Building BrewCo CRM (full-stack CRM, adding an ML churn-prediction layer) and extending SentinelML (fraud detection & explainability)
  • 🌱 Deepening Advanced SQL, Data Warehousing, Deep Learning & NLP
  • 💬 Happy to talk about SQL, data pipelines, FastAPI, or applied ML

Tech Stack

Languages

Data & Analytics

AI & Machine Learning

Backend & Frontend

Databases, Cloud & DevOps


Featured Projects

🤖 SentinelML Fraud detection & explainability system on 284K+ transactions — tuned XGBoost (0.878 PR-AUC, 0.92 precision), PyTorch autoencoder, SHAP explanations, MLflow tracking, deployed as a live API + dashboard. Python XGBoost PyTorch SHAP FastAPI 🔗 Live Demo

Football Analytics Warehouse ETL pipeline transforming 230K+ matches (2000–2025) into a Dockerized PostgreSQL star schema, powering a 12-view analytical layer and a 4-page Power BI dashboard. Python PostgreSQL Docker Power BI ETL

BrewCo CRM Full-stack CRM for a coffee brand — React + FastAPI + PostgreSQL, with AI-driven customer segmentation, campaign management, and a churn-prediction model. Found & patched 2 real PII-leakage bypasses in the AI-generated SQL layer. React FastAPI PostgreSQL Clerk Groq API 🔗 Live Demo

😊 Moodline (Emotion Prediction) NLP emotion classifier using frozen MiniLM embeddings + BiGRU, reaching 86% test accuracy with confidence thresholding to flag uncertain predictions instead of forcing a label. NLP PyTorch MiniLM Classification 🔗 Live Demo


GitHub Stats

GitHub Streak


Let's Connect

Actively looking for opportunities in Data Analytics, Data Engineering, and AI/ML Engineering — open to discussing projects, internships, and collaborations.

profile views

Pinned Loading

  1. brewco-crm brewco-crm Public

    AI-powered, microservice-backed CRM for a fictional coffee brand — RFM customer clustering, churn prediction, natural-language "Ask Your Data" analytics, and AI-assisted campaigns, all behind Clerk…

    JavaScript

  2. EMOTION-PREDICTION EMOTION-PREDICTION Public

    End-to-end Deep Learning application for 6-class emotion classification from text using pretrained MiniLM token embeddings and a BiGRU sequence classifier. The project covers text preprocessing, tr…

    Jupyter Notebook

  3. Football-Analytics-Warehouse Football-Analytics-Warehouse Public

    End-to-end Football Analytics & Data Engineering platform transforming 230K+ football matches (2000–2025) through a Python ETL pipeline into a PostgreSQL star-schema data warehouse, with 12 analyti…

    Python

  4. sentinelml sentinelml Public

    An end-to-end credit card fraud detection and explainability platform featuring multi-model evaluation, Optuna-tuned XGBoost risk scoring, SHAP-based per-transaction attribution, F1-optimized decis…

    Python