Learn AI/ML interactively at AI-ML Companion - Guided walkthroughs, architecture decisions, hands-on challenges, and narrated overviews for every project.
A comprehensive AI/ML learning platform with 10 end-to-end projects covering the full spectrum, from classical ML to production LLM systems. Every project follows industry best-practice structure.
These 10 projects are the hands-on companion to AI-ML Companion, an interactive platform for learning AI and machine learning by watching it work. Instead of passive video, every concept is taught through live, animated visualizations and runnable code: gradient descent stepping down a loss surface, attention weights lighting up across tokens, a network training in real time.
- 22 tracks, 300+ modules from Python and math foundations to deep learning, LLMs, agentic AI, MLOps, and cloud GenAI.
- Interactive visualizations, not slides. See the mechanism move, then experiment with it.
- Runnable Python in the browser, quizzes, and progress tracking.
- Trilingual narration (English, Hindi, Spanish) with technical terms kept in English.
- PWA, works on phone and desktop, installable and offline-capable.
- Free tier: 8 complete tracks (Foundations, Python for ML, Jupyter, Linear Algebra, ML Algorithms, Git, AI Tools, Interview Q&A) plus the first module of every Premium track. Premium unlocks the full catalog.
Curated learning paths: Data Science and AI Engineering guide you end to end.
Explore: Curriculum · Roadmap · Blog
Is AI-ML Companion free? Yes, there is a free tier: 8 complete tracks plus the opening module of every Premium track. Premium unlocks all 22 tracks and 300+ modules.
How is it different from video courses? It is interactive. Concepts are taught through live visualizations and runnable code rather than passive video, so you can watch algorithms operate and experiment directly.
Who is it for? Beginners wanting an intuition-first on-ramp, engineers moving into ML / MLOps / LLM / agentic AI roles, and anyone preparing for AI/ML interviews.
What languages are supported? English, Hindi, and Spanish, with technical terminology kept in English.
| # | Project | Domain | Difficulty | Key Tech | Walkthrough |
|---|---|---|---|---|---|
| 1 | IPL Analysis | Data Science / EDA | Beginner-Intermediate | Pandas, Plotly, Scikit-learn | Learn → |
| 2 | ML Algorithms | Classical ML / Interpretability | Intermediate | Scikit-learn, XGBoost, SHAP | Learn → |
| 3 | Deep Learning | Computer Vision / DL | Intermediate-Advanced | PyTorch, TorchVision | Learn → |
| 4 | ML Pipeline | Feature Engineering / Production ML | Advanced | Scikit-learn, FastAPI, Docker | Learn → |
| 5 | MLOps | Model Deployment / Infrastructure | Advanced | FastAPI, Docker, Prometheus, GitHub Actions | Learn → |
| 6 | LLM/RAG | Retrieval-Augmented Generation | Advanced | LangChain, ChromaDB | Learn → |
| 7 | AI Agents | LLM Agent Orchestration | Advanced | LangGraph, OpenAI, Tavily | Learn → |
| 8 | Content Moderation | Multi-Agentic AI | Advanced | LangGraph, Multi-Agent | Learn → |
| 9 | Due Diligence Agent | Multi-Agent Research | Advanced | LangGraph, Gemini, Streamlit | Learn → |
| 10 | Smart Claims Processor | Multi-Agent Insurance Claims | Advanced | LangGraph, CrewAI, Gemini, FastAPI, React | Learn → |
Comprehensive analysis of 17 IPL seasons with interactive visualizations, hypothesis testing, feature engineering, and predictive modeling.
Highlights: 1000+ matches | Plotly interactive charts | Hypothesis testing | RF + GB models
Compare 6 ML algorithms on real clinical data with cost-sensitive threshold tuning (~95% malignant recall) and SHAP explainability for regulatory review.
Highlights: 6 algorithms compared | XGBoost AUC ~0.994 | SHAP reports | Threshold tuning
Systematically improve a CIFAR-10 image classifier from 60% to 93%+ accuracy across 6 documented experiments with a full diagnostics toolkit.
Highlights: 6 progressive experiments | ResNet + CutMix | LR Finder | Per-class analysis
End-to-end pipeline from messy bank data to deployed model with KNN imputation, domain feature engineering, and PSI drift monitoring.
Highlights: Feature engineering | 10:1 cost-sensitive | PSI drift detection | FastAPI + Docker
Production ML infrastructure: FastAPI with graceful shutdown, CI/CD pipeline, Prometheus metrics, Locust load testing, and operational runbook.
Highlights: CI/CD (GitHub Actions) | P95 < 45ms | 161.7 RPS | Kubernetes-ready
Production RAG system with chunking, security defense, and evaluation framework.
Highlights: RAG pipeline | PII defense | A/B testing
4-agent orchestrated research pipeline (researcher, analyst, writer, fact-checker) with guardrails, evaluation, and cost tracking.
Highlights: LangGraph orchestration | +33% completeness vs single-agent | Budget tracking | LLM-as-judge
Multi-agent content moderation pipeline with specialized agents for different content types.
Enterprise-grade company research powered by 6 AI agents with parallel execution, fact-checking, contradiction resolution, and comprehensive guardrails.
Highlights: 6 specialist agents | Parallel execution via LangGraph Send() | Fact-checking + debate | Streamlit dashboard
Production-style multi-agent insurance claims system built with LangGraph (orchestration) and CrewAI (fraud detection). 7 specialist agents handle intake validation, fraud detection, damage assessment, policy compliance, settlement calculation, LLM-as-judge evaluation, and claimant notification.
Highlights: LangGraph + CrewAI hybrid | Human-in-the-Loop with durable checkpointing | Per-agent confidence gates | Country-aware (US/India) | Pluggable LLMs (Gemini/Groq) | React UI with Agent Trace panel
Every project follows a consistent structure adapted from top ML teams:
project/
├── configs/ # Experiment configuration (YAML)
├── notebooks/ # Exploration & communication
├── src/ # Production source code
├── tests/ # Testing pyramid (unit/integration/load)
├── artifacts/ # Versioned outputs (models, results, figures)
├── docs/ # Model cards, architecture docs, experiment logs
├── scripts/ # One-command automation scripts
├── docker/ # Containerization (where applicable)
├── .gitignore
├── Makefile # make train | make test | make serve
├── requirements.txt
└── README.md
| Principle | What It Means |
|---|---|
| Separation of Concerns | Code (src/), config (configs/), data (data/), and artifacts (artifacts/) never mix |
| Reproducibility First | Configs are YAML, seeds are explicit, environments are containerized |
| Notebook = Communication | Notebooks prototype and communicate; src/ is the production code |
| Testing Pyramid | Unit tests catch logic bugs, integration tests catch pipeline bugs, load tests catch scaling bugs |
| Security by Default | Input sanitization, PII detection, injection defense (critical for LLM projects) |
| Observable from Day 1 | Monitoring, structured logging, metrics export built-in |
Each project is self-contained. Pick one and follow its README:
cd projects/algorithm-showdown # or any other project
pip install -r requirements.txt
make all # train -> evaluate -> test1. IPL Analysis -> Data wrangling, EDA, visualization fundamentals
|
2. ML Algorithms -> Classical ML, model comparison, interpretability
|
3. Deep Learning -> Neural networks, progressive experimentation
|
4. ML Pipeline -> Feature engineering, end-to-end pipelines, monitoring
|
5. MLOps -> Deployment, CI/CD, load testing, infrastructure
|
6. LLM/RAG -> Retrieval-augmented generation, evaluation, security
|
7. AI Agents -> Multi-agent orchestration, guardrails, cost optimization
|
8. Content Moderation -> Multi-agentic content pipelines
|
9. Due Diligence Agent -> Enterprise multi-agent research, fact-checking, debate
|
10. Smart Claims Processor -> Multi-agent insurance, HITL, hybrid orchestration
aiml-companion/
├── projects/
│ ├── ipl-match-predictor/ # EDA + Predictive Modeling
│ ├── algorithm-showdown/ # Classical ML + SHAP
│ ├── deep-learning-project/ # CIFAR-10 + PyTorch
│ ├── credit-risk-pipeline/ # Credit Risk + Monitoring
│ ├── model-serving-platform/ # Model Serving + CI/CD
│ ├── rag-expert-assistant/ # RAG + Security
│ ├── ai-agents-project/ # Multi-Agent + LangGraph
│ ├── content-moderation-project/ # Multi-Agentic Content Moderation
│ ├── due-diligence-agent/ # Multi-Agent Company Research
│ └── smart-claims-processor/ # Multi-Agent Insurance Claims
└── README.md # This file
Author: Rajesh Srivastava