What You'll Build & Deploy
Every project is based on a real system running inside Indian and global companies today — not toy examples. You graduate with a GitHub portfolio that speaks for itself in interviews.
Ship Projects Employers Actually Pay For
Every project is based on a real system running inside Indian and global companies today — not toy examples. You graduate with a GitHub portfolio that speaks for itself in interviews.
Customer Churn Prediction System
End-to-end ML pipeline that predicts which telecom or SaaS customers will cancel their subscription — with business-ready dashboards and alert thresholds.
What you'll build
- Feature engineering pipeline on 15+ customer behavioural signals
- XGBoost + SHAP explainability dashboard so business teams understand why someone will churn
- REST API deployed on Render with real-time scoring & Streamlit monitoring UI
Tech Stack
Used by: Jio, Airtel, Hotstar, SaaS companies — used by growth & retention teams to reduce CAC by 30–60%
Real-Time Fraud Detection Engine
A streaming ML system that flags fraudulent transactions in under 50 ms by combining rule-based filters with an isolation forest model — all in a production-grade pipeline.
What you'll build
- Feature store with velocity features (transactions per hour, geo-distance anomaly)
- Isolation Forest + LightGBM ensemble with <50 ms inference on Kafka stream
- Alert system with Slack notifications and an explainability report per flagged transaction
Tech Stack
Used by: Razorpay, Paytm, HDFC Bank, PhonePe — every payment company runs a system like this 24/7
Stock Sentiment & Price Movement Predictor
Combines NLP sentiment analysis on financial news & Reddit with LSTM time-series forecasting to predict next-day stock price direction for NSE/BSE stocks.
What you'll build
- News scraper + Reddit API pipeline feeding into a FinBERT sentiment classifier
- LSTM model fusing sentiment scores with OHLCV data for directional prediction
- Interactive Plotly Dash dashboard with live portfolio P&L simulation
Tech Stack
Used by: Quantitative desks at NSE brokerages, hedge funds, and algo-trading startups like Zerodha & Samco
AI HR Recruiter Agent
Multi-agent system that screens CVs against a job description, ranks candidates, auto-generates tailored interview question sets, and sends calendar invites — replacing 80% of manual HR work.
What you'll build
- PDF CV parser → vector DB → semantic similarity ranking pipeline
- LangGraph orchestrator with 3 specialised agents (Screener, Ranker, Interviewer)
- Automated email/calendar integration via Gmail API with structured JSON output
Tech Stack
Used by: HR SaaS products like Keka, Darwinbox, Freshteam — and any startup hiring at scale without a large recruiter team
WhatsApp Business AI Assistant
A production-ready WhatsApp bot that handles customer queries 24/7, processes orders, books appointments, and escalates to a human when it detects frustration — all powered by RAG over your business knowledge base.
What you'll build
- Twilio WhatsApp webhook → intent classifier → RAG responder pipeline
- Emotion detection layer that auto-escalates negative-sentiment conversations to a human agent
- Admin dashboard showing conversation logs, escalation rate, and knowledge-base gaps
Tech Stack
Used by: D2C e-commerce brands, clinics, real-estate agencies — any business that gets 100+ repetitive WhatsApp messages daily
Legal Contract Analyser & Risk Flagger
RAG-powered system that ingests contracts (PDF/DOCX), identifies high-risk clauses (indemnity, non-compete, penalties), compares them to standard templates, and answers free-text questions about the document.
What you'll build
- Multi-format document ingestion pipeline (PDF, DOCX, scanned images via OCR)
- Clause-level chunking + custom retriever trained to surface risk-relevant passages first
- Interactive Q&A UI with side-by-side clause comparison and a risk-score heat map
Tech Stack
Used by: Law firms, procurement teams at enterprises, startups reviewing investor term sheets — reduces contract review from days to minutes
AI Job Portal with Smart CV Matching
Full-stack platform where candidates upload their CV and instantly see job recommendations ranked by semantic similarity, plus AI-generated gap analysis and a prep plan for each role.
What you'll build
- Next.js frontend with CV upload → OpenAI Embeddings → Supabase pgvector job search
- Gap analysis agent that compares CV skills to JD requirements and suggests courses
- Admin panel for companies to post jobs and view AI-scored candidate shortlists
Tech Stack
Used by: Naukri, LinkedIn, Instahyre — any job platform where manual screening creates a bottleneck. Also used internally by large enterprises for internal mobility
AI News Aggregator with Personalised Feed
Real-time app that scrapes 50+ news sources, deduplicates stories using NLP, summarises each with GPT-4o-mini, and learns your preferences to serve a ranked personalised feed — like a smart RSS reader.
What you'll build
- Scheduled scraper (Bull queues) + NLP deduplication using cosine similarity on embeddings
- GPT-4o-mini summarisation pipeline with TL;DR, key entities, and sentiment tag per article
- React frontend with user preference learning via implicit feedback (clicks, saves, skips)
Tech Stack
Used by: Media aggregators like Inshorts, Feedly, Briefing — and B2B intelligence products that monitor competitor news for sales teams
End-to-End ML Platform on AWS
Production-grade ML platform with automated retraining pipelines, experiment tracking, model registry, A/B deployment, and rollback — the kind of system that runs ML at every funded startup and large tech company.
What you'll build
- Data pipeline (S3 → feature store → training) orchestrated with Apache Airflow on EC2
- MLflow experiment tracking + model registry with automated evaluation gates
- CI/CD pipeline (GitHub Actions → ECR → SageMaker Endpoint) with blue/green deployment
Tech Stack
Used by: Every company with ML in production: Swiggy, Meesho, CRED, Amazon India — MLOps engineers who build this earn ₹25–50 LPA
LLM Observability & Cost Dashboard
Monitoring system for GenAI applications that tracks token costs, latency by prompt type, hallucination rate (via LLM-as-judge), user satisfaction scores, and alerts the team when quality degrades.
What you'll build
- Langfuse integration to capture every LLM call with latency, tokens, cost, and user feedback
- LLM-as-judge pipeline that evaluates 5% of responses for hallucination and coherence
- Grafana dashboard with cost-per-feature alerts and a weekly model quality email report
Tech Stack
Used by: Any team running GPT-4o at scale — startups spending ₹5L+/month on LLM API costs and enterprises with compliance requirements for AI output quality
What Sets Your Portfolio Apart
Most bootcamps give you notebook exercises. We make you build, deploy, and explain production systems. Hiring managers at top companies have told us: "Your students' projects look like they came from our own engineers."