Advanced Programme in Applied AI & Machine Learning
Course Duration
12 months
Time Commitment
10 Hours/Week
Eligibility
12th Pass and Above
Learning Mode
Online
4.25K+ students have already registered
Course Duration
12 months
Time Commitment
10 Hours/Week
Eligibility
12th Pass and Above
Learning Mode
Online
Why Choose This Course?
Learn Like an IITian, Excel Like a Global Leader
Receive a Certificate of Completion with Vishlesan i-Hub, IIT Patna, significantly enhancing your profile's value.
Work on capstone projects designed to solve practical business challenges.
Learn from IIT faculty & industry experts bringing real-world insights to your learning journey
Receive resume reviews, career coaching, and placement support to land your dream role. (Eligibility: 7+ CGPA and 65% attendance)
Prestigious Certification*:
Receive a Certificate of Completion with Vishlesan i-Hub, IIT Patna, significantly enhancing your profile's value.
Real-World Capstone Project:
Work on capstone projects designed to solve practical business challenges.
Expert Faculty & Industry Insights:
Learn from IIT faculty & industry experts bringing real-world insights to your learning journey
Placement Opportunities
Receive resume reviews, career coaching, and placement support to land your dream role. (Eligibility: 7+ CGPA and 65% attendance)
*Certificate Eligibility: Minimum 65% overall attendance and 30% marks in cumulative evaluations score
What Will You Learn?
Begin your AI/ML journey with a beginner-friendly, industry-aligned curriculum from Vishlesan i-Hub, IIT Patna. Covering everything from Python basics to advanced topics like deep learning and generative AI, the program builds real-world skills through hands-on projects and expert-led learning.
Toolkit
Course Details
Duration
12 Months
Course Mode
Online
Certification
from Vishlesan i-Hub IIT Patna
- • Build working Python fluency for AI/ML — functions, file handling, APIs, clean coding habits, and modular project structuring
- • Develop professional EDA workflows: hypothesis-driven exploration, data profiling, missing-value strategies, outlier treatment, feature diagnostics, leakage detection, and structured EDA reporting
- • Go deep on the mathematics that powers every model you will ever build — matrices, vector spaces, calculus for optimisation, probability distributions, and statistical inference
- • Build intuition for entropy, gradients, and regularisation — the concepts that separate practitioners who debug models from those who merely run them
- • Establish version control, notebook discipline, and reproducible project structure as non-negotiable engineering defaults from week one
- • Achieve deep mastery across every ML family — regression, classification, clustering, tree-based, and distance-based models — with full coverage of model assumptions and when they break
- • Build strong baselines before iterating: understand overfitting, underfitting, bias-variance tradeoff, and regularisation as diagnostic tools, not afterthoughts
- • Evaluate with rigour — precision, recall, AUC, calibration, and metric selection — and understand why the wrong metric can make a failing model look like a success
- • Build end-to-end feature pipelines with cross-validation, hyperparameter tuning, experiment tracking, and SHAP-based model interpretation at production scale
- • Produce reproducible, documented ML workflows that another engineer can inherit, audit, and extend — not just notebooks that ran once
- • Build and train deep networks from first principles — MLPs, backpropagation, optimisers, regularisation, and training diagnostics that actually tell you what is going wrong
- • Master CNNs and RNNs with full transfer learning workflows, and develop attention mechanism intuition and a working understanding of transformer architecture
- • Go deeper on NLP — tokenisation, embeddings, transformer-based NLP models, and deployment-ready text classification and extraction pipelines
- • Build vision systems with CNNs, pretrained models, and multimodal model intuition for understanding how text and image understanding converge
- • Evaluate every model with deployment in mind — not just accuracy, but latency, size, and failure mode analysis for real-world use
- • Develop deep architectural understanding of LLMs — transformers, self-attention, positional encoding, pretraining, RLHF, context windows, and inference behaviour at scale
- • Design production-grade LLM application flows: prompt architecture, system prompts, structured JSON outputs, function calling, and tool orchestration
- • Build and evaluate LLM pipelines with prompt testing frameworks, eval datasets, safety boundaries, and user experience design built in from the start
- • Understand how model behaviour shifts with scale, instruction tuning, and context design — so you know what you are working with before you build on top of it
- • Go beyond single-turn prompting — design multi-turn flows, tool-augmented systems, and LLM applications that hold up under real user behaviour
- • Build production-grade RAG with advanced chunking strategies, metadata filtering, hybrid retrieval, reranking, and query rewriting — not toy demos
- • Evaluate RAG pipelines rigorously: citation accuracy, retrieval quality, failure analysis, latency benchmarking, and cost modelling at scale
- • Go deep on fine-tuning: SFT, LoRA, and QLoRA workflows with dataset curation, tokenisation, training configs, and overfitting risk management
- • Compare fine-tuned models against prompted and RAG-based baselines with structured evaluation — and know which approach to defend and why
- • Understand deployment considerations for adapted models: size, serving cost, versioning, and when a fine-tuned model creates more problems than it solves
- • Build advanced agentic systems using graph-based architectures and state machines — far beyond simple tool-calling chatbots
- • Design planning loops, conditional routing, and multi-step task execution with stateful memory that persists across interactions
- • Implement multi-agent patterns: orchestrator-worker architectures, agent handoffs, shared memory, and conflict resolution between agents
- • Build in failure handling, observability, and reliability testing — because agents that fail silently in production are more dangerous than agents that do not ship
- • Evaluate agent performance with structured benchmarks: task completion rate, tool call efficiency, hallucination rate, and cost per resolved task
- • Deploy ML models and GenAI apps as live, versioned, monitored REST APIs — FastAPI, Docker, cloud deployment, and structured logging built in from day one
- • Build LLMOps infrastructure: prompt versioning, tracing, eval sets, cost tracking, latency monitoring, and user feedback loops that actually improve the system over time
- • Implement guardrails, fallback behaviour, and model update strategies that make production GenAI apps safe, measurable, and maintainable
- • Apply senior engineering workflow to every project — branching strategy, pull requests, environment files, reproducibility standards, and team-ready documentation
- • Graduate with a portfolio of production-deployed systems — not notebooks, not demos, but live APIs with monitoring, versioning, and a deployment story you can walk any engineering team through
Please Note :- "Due to the evolving nature of the industry expectations and partner institute feedback, some syllabus aspects may change. Any updates will be communicated during the Inauguration Session(s) or at the start of the relevant module". The project deliverables or case studies for your batch may differ, as per the decision of the Institute or Course Faculty
Meet Our Instructor Team
*Note: Instructors teaching your batch may change based on specific module expertise, or circumstances such as personal emergencies, unavailability etc.