# Welcome! You already ship software: pipelines, environments, rollouts, observability. **Machine learning is that same muscle** — except the "binary" you deploy is a model, and the bugs can drift in silently like a change in crowd noise at Wankhede. This repo is your **two-track studio album**: fast interview-ready takes, and deeper "liner notes" when you want the full mix. *** ## Repo Structure ``` 01-foundations/ Intro to AI, glossary, revision guide 02-classical-ml/ Supervised, unsupervised, preprocessing, anomaly detection, active learning, Bayesian methods 03-deep-learning/ Components + methods + transfer learning + video + 3D vision + PyTorch 04-specialized-domains/ RL, RecSys, GNNs 05-llms/ Architecture, training, applications, interview notes 06-production-ml/ MLOps, system design 07-interview-prep/ ML + LLM interview Q&As, scenarios, coding 08-emerging-topics/ Emerging trends, XAI, causal inference, privacy-preserving ML, continual learning 09-study-plans/ 30-day week-by-week study plan 10-references/ Book notes, research papers, datasets, tools, benchmarks, notation MIND-MAP.md Full topic mind map + gap analysis ``` *** ## Track 1 — Interview Blueprint (high tempo) **Where to start tonight** * [**AI & ML revision guide**](/artificial-intelligence/ai_ml_revision_guide.md) — Golden rules, architectures, math you can whiteboard, infra patterns, gotchas. * [**30-day roadmap**](/artificial-intelligence/09-study-plans/09-study-plans.md) — Structured reps from zero to "I can hold the room." * [**Math derivations hub**](/artificial-intelligence/ml-interview-notes/math-derivations.md) — Chain rule to attention, with your pen. * [**LLM interview notes**](/artificial-intelligence/interview-notes/interview-notes.md) — RAG, agents, alignment, serving — the GenAI setlist. * [**ML interview notes**](/artificial-intelligence/ml-interview-notes/ml.md) — Full classical + DL Q\&A bank. * [**Pre-interview checklist**](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/07-interview-prep/PRE-INTERVIEW-CHECKLIST.md) — 48h → 24h → morning-of: what to review, verify, and do before walking in. **10-Minute Revision Cards** (skim before any topic or interview) | Section | Revision Card | | ------------------------------------- | ------------------------------------------------------------------------------------------------------------------------ | | Foundations | [REVISION.md](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/01-foundations/REVISION.md) | | Classical ML | [REVISION.md](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/02-classical-ml/REVISION.md) | | Deep Learning | [REVISION.md](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/03-deep-learning/REVISION.md) | | Specialized Domains (RL, RecSys, GNN) | [REVISION.md](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/04-specialized-domains/REVISION.md) | | LLMs | [REVISION.md](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/05-llms/REVISION.md) | | Production ML | [REVISION.md](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/06-production-ml/REVISION.md) | **Azure / DevOps bridge:** *Training is your build job; the model artifact is your release candidate; inference is the always-on service; MLOps is CI/CD when the "code" and the "data" both change.* *** ## Track 2 — Deep-Dive Library (studio sessions) | Section | What's inside | | ----------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------- | | [Foundations](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/01-foundations/README.md) | Intro to AI, glossary, revision guide | | [Classical ML](/artificial-intelligence/02-classical-ml/02-classical-ml.md) | Bias–variance, trees, calibration, when classical beats DL | | [Deep Learning](/artificial-intelligence/03-deep-learning/03-deep-learning.md) | Activations, backprop, attention, PyTorch, MCP | | [Computer Vision](/artificial-intelligence/methods/computer-vision.md) | CNNs, detection, ViT, CLIP, self-supervised | | [Generative Models](/artificial-intelligence/methods/generative-models.md) | VAE, GAN, DDPM/DDIM, Stable Diffusion, Flow Matching, DiT/FLUX | | [Time Series](/artificial-intelligence/methods/time-series.md) | ARIMA through Transformers, forecasting, anomaly detection | | [Reinforcement Learning](/artificial-intelligence/04-specialized-domains/reinforcement-learning.md) | MDPs, Q-learning, PPO, RLHF connection | | [Recommender Systems](/artificial-intelligence/04-specialized-domains/recommender-systems.md) | Collaborative filtering, two-tower, ranking | | [Graph Neural Networks](/artificial-intelligence/04-specialized-domains/graph-neural-networks.md) | GCN, GraphSAGE, knowledge graphs | | [LLMs](/artificial-intelligence/05-llms/05-llms.md) | Architecture, training, scaling, evaluation | | [LLM Applications](/artificial-intelligence/applications/applications.md) | RAG, agents, tuning, inference optimization | | [Multimodal AI](/artificial-intelligence/applications/multimodal.md) | CLIP, VLMs, fusion architectures, audio, video, deployment | | [Speculative Decoding](/artificial-intelligence/applications/speculative-decoding.md) | Medusa, Eagle, standard SD, production trade-offs | | [LLM Training Stability](/artificial-intelligence/05-llms/training-stability.md) | Loss spikes, mixed precision, RLHF failure modes | | [Production ML](/artificial-intelligence/06-production-ml/06-production-ml.md) | MLOps, CI/CD for ML, deployment | | [Model Governance](/artificial-intelligence/06-production-ml/model-governance.md) | Model registry, audit trails, GDPR, champion-challenger | | [ML System Design](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/06-production-ml/system-design/README.md) | Design patterns, case studies, engineering | | [Interpretability & XAI](/artificial-intelligence/08-emerging-topics/interpretability-and-xai.md) | SHAP, LIME, mechanistic interpretability | | [Causal Inference](/artificial-intelligence/08-emerging-topics/experimentation-and-causal-inference.md) | A/B testing, causal graphs, uplift modeling | | [Emerging Trends 2023–2025](/artificial-intelligence/08-emerging-topics/emerging-trends.md) | Mamba, MoE, test-time scaling, synthetic data, long context | | [2025 Frontier Models](/artificial-intelligence/08-emerging-topics/2025-frontier-models.md) | DeepSeek, Llama 4, Gemini 2.5, Claude 3.7, GPT-o3, Qwen3 | | [Model Compression](/artificial-intelligence/components/model-compression.md) | Quantization, distillation, pruning | | [Research Papers](/artificial-intelligence/research-papers/research-papers.md) | 25 foundational LLM papers every ML interview cares about | | [Anomaly Detection](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/02-classical-ml/anomaly-detection.md) | IForest, OCSVM, LOF, Autoencoders | | [Active Learning](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/02-classical-ml/active-learning.md) | Uncertainty sampling, QbC, Core-Set, BADGE | | [Bayesian Methods](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/02-classical-ml/bayesian-methods.md) | GPs, BNNs, variational inference, Bayesian optimization | | [Conformal Prediction](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/02-classical-ml/conformal-prediction.md) | Distribution-free coverage guarantees, CQR | | [Transfer Learning & Domain Adaptation](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/03-deep-learning/transfer-learning.md) | Fine-tuning, DANN, few-shot, MAML, zero-shot | | [Video Understanding](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/03-deep-learning/methods/video-understanding.md) | Two-stream, I3D, SlowFast, Video Transformers | | [3D Vision & Point Clouds](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/03-deep-learning/methods/3d-vision.md) | PointNet, PointNet++, NeRF, 3DGS, autonomous driving | | [Privacy-Preserving ML](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/08-emerging-topics/privacy-preserving-ml.md) | DP-SGD, Federated Learning, SMPC, HE | | [Continual Learning & NAS](https://github.com/nishchalnishant/Ml-overview/blob/modifying-using-ai/08-emerging-topics/continual-learning.md) | Catastrophic forgetting, EWC, replay, DARTS | *** ## How These Notes Are Written Each topic aims for a **senior answer in three beats**: 1. **Direct line** — What you say in the first ten seconds. 2. **Intuition** — An analogy that sticks. 3. **Production** — Latency, cost, scale, and what breaks first. *** > **Cold open:** Production ML is a lot of engineering with a little bit of "magic." This repo leans into the engineering — because that's what keeps models out of the **Incident** channel. --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://nishchalnishant.gitbook.io/artificial-intelligence/readme.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.