2+ yrs production experience (TCS) · Ex JP Morgan · Ex IIT Indore · 🥇 Google Challenge 2026 Winner

I don't just build models — I ship them. I've spent 2+ years at Tata Consultancy Services building production data systems across the technology and financial sectors. I containerize, deploy, monitor, and iterate. If it doesn't run in production, I'm not done.

Currently completing my MSc in Data Science at the University of Naples Federico II, where I led a team to 1st Place (STEM) at Google Challenge Campania 2026 — building a generative AI solution that beat out every competing team.

• Full-lifecycle ML: data engineering → model development → containerized deployment → monitoring
• Production systems: FastAPI + Docker + AWS — not just notebooks
• Big data at scale: PySpark, Kafka, Spark NLP on high-volume streaming data
• Real business impact: 2+ years solving problems in technology and financial services at TCS & JP Morgan

Domain	Technologies
ML & Deep Learning
Big Data & Streaming
MLOps & Cloud
Languages & Data

Every project below is deployed, documented, or has a live demo. Click through.

What: Production REST API that classifies network traffic as phishing or legitimate — in real time and in batch.

Impact: Deployed and running on AWS EC2 with zero-downtime containerized architecture. Serves predictions via FastAPI endpoints with full experiment tracking.

How it works: Raw network data → feature extraction → trained ML model → prediction served via REST endpoint → results logged in MongoDB → experiments versioned in MLflow.

FastAPI Docker AWS EC2 MLflow MongoDB Scikit-learn

What: Self-hosted AI agent that discovers job postings, scores them against your profile, and tailors applications — automatically.

Impact: Runs 24/7 in Docker for ~$1.80/month. Replaces hours of manual job hunting with intelligent, automated matching.

How it works: n8n orchestration → job discovery → Claude API scores & matches → PostgreSQL stores state → tailored applications generated automatically.

n8n Claude API PostgreSQL Docker Shell

What: Deep learning model that predicts which customers will leave — before they do.

Impact: ANN with dropout regularization for production-grade binary classification. Live Streamlit app lets anyone input customer data and get instant predictions.

TensorFlow Keras Streamlit Deep Learning

What: Hybrid recommender combining content-based filtering and collaborative filtering.

Impact: Deployed interactive app that returns instant movie recommendations using cosine similarity on movie metadata.

Scikit-learn NLP Streamlit Python

What: Distributed pipeline for real-time sentiment and fault classification on high-volume data streams.

Impact: Handles streaming data at scale using PySpark + Kafka, with NLP classification applied in real time — not batch.

PySpark Kafka Spark NLP Big Data

What: Classification pipeline predicting whether a customer will subscribe to a term deposit.

Impact: Full ML pipeline with PCA dimensionality reduction and GridSearchCV hyperparameter tuning. Evaluated on ROC-AUC for real-world class imbalance.

Scikit-learn ROC-AUC GridSearchCV Random Forest

Open to opportunities in Data Science, ML Engineering & MLOps
LinkedIn · Portfolio · deepakkushwaha771@gmail.com