Overview
An interview kit for data scientists moving into GenAI roles. It covers the full GenAI engineering lifecycle, from prompting and RAG to agents, fine-tuning, evaluation, deployment and monitoring, plus application engineering, transformer internals and multimodal systems, while keeping the data science strengths interviewers look for: statistics, experimentation, metrics, data quality and business framing. Sixteen volumes each have topic sections, a hands-on section with broken prompts, traces, SQL, code and configs to diagnose, and a scenario section. Many answers carry a diagram.
What you will learn
- Bridge classical data science and GenAI in interview answers
- Use LLMs and embeddings as data science tools
- Build and evaluate RAG systems and analytics copilots
- Fine-tune models with a data-centric approach
- Evaluate LLMs with statistical rigor and run A/B tests on GenAI features
- Deploy, monitor and cost GenAI systems responsibly
- Design hybrid ML and GenAI systems
- Explain transformer internals and multimodal models
Included with the kit
- 100 written parts, yours for good
- 28h 21m of reading, measured not estimated
- Written for the intermediate level
- Every future revision included
- UPI, cards and netbanking
Prerequisites
- Data science experience: Python, statistics and machine learning
- No prior GenAI production experience needed
Curriculum
16 sections · 100 parts · 28h 21m01Volume 1: Data Science Foundations Refresher for GenAI6 parts · 101 min
Show that your statistics, classical ML, data wrangling, and problem framing skills carry over to GenAI work, and explain clearly where they need to change.
- Statistics & Probability in GenAI Work16 min
- Classical ML & Metrics Carried into GenAI17 min
- Data Wrangling, SQL & Data Quality for LLM Projects17 min
- Framing Business Problems: Classical DS vs GenAI13 min
- DS foundations refresher: hands-on12 min
- DS foundations refresher: scenarios26 min
02Volume 2: GenAI & LLM Foundations for Data Scientists7 parts · 121 min
Build a working understanding of how LLMs process text, what drives their cost and behavior, and how to choose a model, explained from a data scientist's point of view.
- Transformers & Attention for Data Scientists18 min
- Tokens, Context Windows & Cost17 min
- Embeddings & Representations12 min
- Decoding, Sampling & Non-Determinism16 min
- Model Landscape & Model Selection17 min
- GenAI foundations: hands-on12 min
- GenAI foundations: scenarios29 min
03Volume 3: Prompt Engineering for Data Work6 parts · 94 min
Write, structure, test, and version prompts that turn LLMs into reliable tools for labeling, extraction, summarization, and analysis.
- Prompt Design Fundamentals20 min
- Few-Shot, Chain-of-Thought & Reasoning Prompts12 min
- Structured Outputs & Schema Control15 min
- Prompt Testing, Versioning & Iteration13 min
- Prompt engineering for data: hands-on13 min
- Prompt engineering for data: scenarios21 min
04Volume 4: LLMs as Data Science Tools6 parts · 96 min
Use prompted LLMs as practical data science tools for labeling, extraction, entity resolution, EDA, and feature engineering, and know when a classical model is the better choice.
- Text Classification & Labeling with LLMs21 min
- Information Extraction & Entity Resolution11 min
- LLM vs Classical NLP and ML Decisions16 min
- LLM-Assisted EDA, Data Cleaning & Feature Engineering15 min
- LLMs as DS tools: hands-on12 min
- LLMs as DS tools: scenarios21 min
05Volume 5: Embeddings & Semantic Data Science6 parts · 97 min
Choose, evaluate, and apply embeddings for clustering, topic discovery, similarity search, and anomaly detection, and keep them reliable as data and models change.
- Embedding Models & Selection16 min
- Clustering, Topic Modeling & Visualization16 min
- Similarity Search & Vector Databases15 min
- Embedding Quality, Drift & Domain Adaptation15 min
- Embeddings: hands-on11 min
- Embeddings: scenarios24 min
06Volume 6: RAG Systems End to End7 parts · 121 min
Prepare the candidate to design, debug, and defend a full retrieval-augmented generation pipeline, from document ingestion through retrieval, grounded generation, advanced patterns, and structured data, using the measurement habits of a data scientist.
- Ingestion, Parsing & Chunking19 min
- Retrieval, Hybrid Search & Re-Ranking18 min
- Generation, Grounding & Citations16 min
- Advanced RAG Patterns19 min
- RAG over Structured & Tabular Data11 min
- RAG end to end: hands-on12 min
- RAG end to end: scenarios26 min
07Volume 7: Agents & Analytics Copilots6 parts · 106 min
Prepare the candidate to design, build, and safely run tool-using agents and data copilots, including text-to-SQL, multi-agent orchestration, and human-in-the-loop controls, with the rigor a data scientist brings to numbers that business users will act on.
- Agent Fundamentals & Tool Calling14 min
- Text-to-SQL & Data Agents18 min
- Multi-Agent Systems & Orchestration17 min
- Agent Reliability, Guardrails & Human-in-the-Loop16 min
- Agents and copilots: hands-on12 min
- Agents and copilots: scenarios29 min
08Volume 8: Fine-Tuning & Data-Centric GenAI6 parts · 110 min
Decide when fine-tuning is worth it, build and curate the training data that makes it work, run PEFT and alignment methods with sound mechanics, and distill large models into smaller or classical ones.
- When to Fine-Tune vs Prompt vs RAG16 min
- Training Data Curation & Synthetic Data19 min
- PEFT, LoRA & Training Mechanics19 min
- Distillation, Small Models & Alignment18 min
- Fine-tuning: hands-on13 min
- Fine-tuning: scenarios25 min
09Volume 9: LLM Evaluation with Statistical Rigor7 parts · 127 min
Prepare the candidate to design, run, and defend LLM, RAG, and agent evaluations with the statistical rigor a data scientist brings, from golden datasets and judge calibration to confidence intervals and error analysis.
- Evaluation Design & Golden Datasets18 min
- Metrics for Generation, RAG & Agents20 min
- LLM-as-a-Judge & Human Evaluation19 min
- Statistical Significance, Uncertainty & Sample Size17 min
- Error Analysis & Failure Taxonomies11 min
- LLM evaluation: hands-on12 min
- LLM evaluation: scenarios30 min
10Volume 10: Experimentation & A/B Testing for GenAI6 parts · 106 min
Prepare the candidate to design, analyze, and defend online experiments and causal impact studies for GenAI features, handling the variance, interference, and measurement problems that LLM products create.
- Experiment Design for GenAI Features22 min
- Metrics, Guardrails & Proxy Measures14 min
- Analysis Pitfalls: Novelty, Variance & Interference19 min
- Causal Inference & Impact Measurement15 min
- Experimentation and A/B testing: hands-on11 min
- Experimentation and A/B testing: scenarios25 min
11Volume 11: Deployment, LLMOps, Monitoring & Observability7 parts · 111 min
Prepare the candidate to ship, version, monitor, trace, and continuously improve LLM systems in production, using the measurement discipline of a data scientist.
- Serving, Latency & Scaling15 min
- LLMOps Pipelines, Versioning & CI/CD11 min
- Monitoring, Drift & Quality in Production18 min
- Tracing, Observability & Debugging16 min
- Feedback Loops & Continuous Improvement11 min
- Deployment and monitoring: hands-on13 min
- Deployment and monitoring: scenarios27 min
12Volume 12: Cost, Safety, Responsible AI & Business Framing6 parts · 103 min
Prepare the candidate to model and control GenAI costs, secure LLM systems and data, build fair and responsible features, and make an honest business case to stakeholders.
- Cost Modeling & Optimization15 min
- Security, Prompt Injection & Privacy24 min
- Bias, Fairness & Responsible AI19 min
- ROI, Stakeholders & Communicating Uncertainty10 min
- Cost, safety and business: hands-on13 min
- Cost, safety and business: scenarios22 min
13Volume 13: Hybrid ML + GenAI System Design & Transition Interviews6 parts · 110 min
Prepare the candidate to design systems that combine classical ML with LLMs, work through domain case studies, and tell a convincing story about moving from data science into GenAI.
- Hybrid Classical ML + LLM Architectures19 min
- End-to-End System Design16 min
- Domain Case Studies17 min
- Transition Stories, Portfolio & Behavioral Rounds15 min
- Hybrid system design: hands-on13 min
- Hybrid system design: scenarios30 min
14Volume 14: AI Application Engineering & Coding6 parts · 112 min
Prepare a data scientist whose code has lived in notebooks to write, test, package, and ship production-grade LLM application code, from async clients and retries to typed schemas, tests, containers, secrets, and API auth.
- Python LLM Clients, Async & Streaming16 min
- Timeouts, Retries, Rate Limits & Queues16 min
- Typed Schemas, Validation & Testing17 min
- Packaging, Deployment, Secrets & Auth14 min
- AI application engineering: hands-on28 min
- AI application engineering: scenarios21 min
15Volume 15: Transformer & Deep Learning Internals6 parts · 101 min
Go below the API and explain how a transformer computes, trains, and serves tokens, including the memory math, PyTorch training mechanics, and inference optimizations that technical GenAI interviews probe.
- Transformer Block Walkthrough16 min
- Training Objectives & the Model Pipeline14 min
- Inference Mechanics: Prefill, Decode, KV Cache & Memory Math17 min
- PyTorch Training Loops, Mixed Precision & Distributed Training15 min
- Transformer internals: hands-on18 min
- Transformer internals: scenarios21 min
16Volume 16: Multimodal GenAI6 parts · 85 min
Build, evaluate, and cost GenAI systems that work with scanned documents, images, audio, and cross-modal retrieval, with modality-specific metrics and grounding a reviewer can check.
- Document AI: OCR, Layout & Tables16 min
- Vision-Language Models & Images14 min
- Audio & Speech11 min
- Multimodal Embeddings, Retrieval & Grounding12 min
- Multimodal GenAI: hands-on13 min
- Multimodal GenAI: scenarios19 min
Reviews
to review this kit once you have finished it.