This isn't a guided tutorial with a predetermined outcome. You choose a real problem, design an AI solution, build it from scratch, deploy it live, and present it to the batch — and to hiring partners. Every decision is yours. Every line of code is yours. Every result is yours.
Scope & Design
Define your problem, validate the approach, and design the system architecture with mentor feedback.
Build & Iterate
Implement core ML/AI components with weekly code reviews. Break things, learn why, fix them better.
Deploy & Polish
Ship to production. Build a live demo. Add monitoring. Make it something you'd show a hiring manager.
Demo Day
Live presentation to the cohort and invited hiring partners. Real feedback. Real opportunity.
Every project is unique to the student's domain and goal. These are real capstones built by recent Newton JEE graduates — all are live in production or on GitHub.
Clinical Note Summariser — RAG pipeline for Indian hospital discharge notes
Built a BERT-based NER model to extract diagnoses and medications from unstructured Hindi-English mixed clinical notes, then wired it to a LangChain RAG pipeline for structured Q&A. Deployed on AWS EC2 with a FastAPI backend serving a Streamlit UI. Over 500 test documents processed. Now piloting with a Hyderabad clinic.
Mutual Fund Comparator with ML-driven risk scoring
Scraped 3 years of NAV data, engineered 40+ features, trained an XGBoost classifier to predict risk-adjusted return quartiles. Deployed as a public API with a React dashboard.
Visual product search for clothing catalogues
CLIP embeddings for image-to-image search, indexed with FAISS. Handles 10K product images with sub-100ms retrieval. Chrome extension as the UI. 800+ GitHub stars.
Personalised Doubt-Solver Agent for JEE Physics
Multi-step reasoning agent with GPT-4, tool calling to a physics formula DB, and step-by-step LaTeX solution rendering. Fine-tuned on 2K JEE solved examples using QLoRA.
Every week, you share your latest code, demo, or architecture doc. Your mentor does a live review — specific, constructive, and laser-focused on making your project better. This is what professional code review feels like.
This is what every review session looks like. You bring a specific problem. Your mentor brings a decade of production experience.
- 1:1 project consultation — choosing a problem you'll care about for 8 weeks
- From idea to ML problem statement: defining inputs, outputs, and success metrics
- Baseline research: what exists, what doesn't, what your project uniquely adds
- Data availability audit — do you have enough of the right kind of data?
- Deliverable: Project Brief (1-page doc with problem, approach, and timeline)
- Designing your full system: components, data flow, model interfaces
- Data collection, cleaning, and versioning with DVC
- Setting up your repo structure the production-engineering way
- Architecture review with mentor — identifying risks before you build
- Deliverable: System design doc + clean data pipeline running
- Rapid prototyping: build v0 by end of week 3, no matter how rough
- Experiment tracking with MLflow — logging every run, metric, and parameter
- Model selection: picking the right approach for your specific problem constraints
- Baseline → improved model → evaluating the delta honestly
- Weekly code review: mentor reads your actual code, not just outputs
- Deliverable: Working model v1 with evaluation metrics documented
- Building a FastAPI or Flask backend to serve your model predictions
- Input validation, error handling, and graceful failure modes
- Logging inference requests for monitoring and debugging
- Frontend or demo UI — Streamlit, Gradio, or basic HTML/React depending on project
- Deliverable: Local demo that a non-technical person can use
- Containerising your app with Docker: reproducibility and clean dependencies
- Deploying to a live URL: AWS EC2, Railway, Render, or Hugging Face Spaces
- Basic monitoring: request logs, latency tracking, alert conditions
- Performance optimisation: reducing latency, memory usage, and cost per prediction
- Writing a strong README and GitHub repo that impresses recruiters
- Deliverable: Live deployed project with public URL
- How to demo an ML project: the 5-minute structure that works in any interview
- Handling hard questions: "Why this model?", "What would you do differently?", "Can it scale?"
- Live demo to cohort — peer feedback in real-time
- Presentation to invited hiring partners from the Newton JEE network
- Platinum certificate issuance + LinkedIn badge activation
- Deliverable: Demo Day presentation (recorded and added to your portfolio)
A Real Portfolio Project
Not a tutorial. Not a Kaggle fork. A system you designed, built, deployed, and can explain from first principles in any interview.
Platinum LinkedIn Badge
The highest certificate in the Newton JEE path. Signals to recruiters that you've gone end-to-end — from raw data to production system.
Live Demo Day Exposure
Your project is presented to actual hiring partners from our network. Some graduates have received interview calls directly from Demo Day.
1:1 Career Coaching
A dedicated session with a mentor to map your next 6 months: which roles to target, how to position your portfolio, which skills to build next.
// This Course Is Right For You If
The Capstone was the hardest and most valuable thing I've done in my career. Meera's weekly reviews were surgical — she didn't tell me what to build, but she asked the right questions until I figured out why my architecture was wrong. My project is now the first thing I show in any interview. I got 4 offers from it.
The Demo Day was surreal. I presented my clinical NLP project to 3 companies and got an interview from one of them the next week. That doesn't happen with any other course. You're not just learning — you're being introduced to the industry.
I'd done 6 courses before the Capstone. The Capstone is where I finally understood that knowing ML and knowing how to build ML systems are completely different skills. The deployment week alone taught me more than 3 months of Udemy courses.
The 1:1 career session with Meera changed my job search strategy completely. She looked at my project, my target companies, and my resume, and gave me a 3-month plan that was more specific and useful than 6 months of general career advice. Got placed within 8 weeks.