Internship programme
AI & Machine Learning Internship
Train it, evaluate it honestly, then actually deploy it.
Plenty of people can fine-tune a model in a notebook. Far fewer can say what it costs to serve, how it fails, or whether it beats the simple thing it replaced. This programme spends as much time on evaluation and deployment as it does on training, because that is where projects die.
Day to day
What you actually do.
- Build a baseline first, then earn the right to use a model
- Train, evaluate and error-analyse on a real dataset
- Ship an LLM application with retrieval and guardrails
- Deploy a model behind an API with latency and cost budgets
- Write a model card documenting limits and failure modes
Structure
12 weeks, phase by phase.
- Weeks 1–3
Data and classical ML
- Python for ML: NumPy, pandas and scikit-learn
- Feature engineering, leakage and why your accuracy was fake
- Regression, trees and gradient boosting
- Cross-validation, metric choice and error analysis
- Weeks 4–6
Deep learning
- Neural networks and training dynamics in PyTorch
- CNNs for vision, transfer learning and augmentation
- Sequence models and the transformer architecture
- Reading a paper and reproducing one result from it
- Weeks 7–9
LLM applications
- Prompt design, structured outputs and function calling
- Retrieval-augmented generation over your own corpus
- Evaluation sets, regression testing and guardrails
- Cost, latency and caching in a real application
- Weeks 10–12
Production
- Serving a model with FastAPI and containers
- Monitoring, drift detection and retraining triggers
- A capstone project deployed and documented
- Portfolio review and interview preparation
What you will use
What you leave with
- A deployed ML or LLM application with a public demo
- An evaluation report and model card in your portfolio
- AICTE-recognised completion certificate
- Mentor reference and interview preparation
Who this is for
Eligibility, stated
plainly.
We would rather turn someone away at the application stage than three weeks into a cohort. If you are unsure whether you qualify, apply and ask — we answer honestly.
- Comfortable writing Python without a tutorial open
- Basic linear algebra, probability and calculus
- Students or graduates in CS, ECE, statistics or mathematics
- Around 20 hours a week
Questions
About this programme.
Do I need a GPU?
No. The programme is built around free and low-cost cloud compute, and we provide credits for the capstone. Anything that genuinely needs a large GPU runs on our infrastructure.
How much mathematics do I need?
Enough to not be frightened by a matrix multiplication and a gradient. We revise what is needed as it comes up rather than front-loading a theory course you would forget.
Other programmes
12 weeks
Full Stack Development
Build and ship a real application, reviewed line by line.
10 weeksFinance & Statistics
Statistics applied to money, with datasets that are not clean.
10 weeksStatistics under a CA
Where audit judgement meets statistical method.
12 weeksIoT & Embedded Systems
Real boards, real firmware, real things that stop working.
Applications
Ready for the ai & machine learning track?
Send us a CV.
One paragraph about what you want to build is worth more to us than a cover letter template. We read every application and reply to every one.
Email hackedinsolutions@gmail.com · Reply within one working day