The Deployed Data Scientist: MLOps and Analytics in Practice, by Ankit Anand, Scott Burk, and Kinshuk Dutta
Transform your Machine Learning Operations (MLOps) projects into reliable and scalable data products that meet the complex demands of data science.
Field Story: The Ghost in the Join
Case Study: The Silent Model Failure
From Projects to Products
The Netflix Recommendation Engine
Illustrative Code Examples
A Bank’s Real-Time Fraud Detection System
The Deployed Data Scientist’s Manifesto
The Data Scientist as a Business Leader
The Ethics of Deployment: A Proactive Responsibility
Chapter Summary
Beyond the CSV
The Modern Data Stack: The Best of All Worlds
Towards Resilient Data Products
The Philosophy of Model Decay: Why All Models Are Born to Fail
Field Story: When Seasons Change
The Necessary Guardrails
Building a Data-First Culture
Chapter Summary
Strategic Model Selection
The Model Selection Decision Framework
Advanced Evaluation Metrics
For Probabilistic Predictions: Log-Loss
The Model Registry: Your Single Source of Truth
Model Interpretability and Explainability
A Preview of Common Techniques
Comprehensive Explainability Framework
Field Story: The Hyperparameter Nightmare
The Silent Failure
Chapter Summary
The Role of Containerization: Docker and Kubernetes
The ML Testing Pyramid: A Stratified Defense
Layer 2: Integration Tests (The Glue)
Layer 3: End-to-End Tests (The Dress Rehearsal)
Continuous Integration (CI): Your Automated Quality Gates
Continuous Delivery (CD): Your Automated Release Engine
Stories from the Assembly Line
Chapter Summary
Architecting a Cloud Strategy: Choosing Your Foundation
The Two Deployment Philosophies: Managed vs. Self-Managed
Architecting for Performance and Cost: Serverless vs. Dedicated
Field Story: The Billion-Request Challenge
Chapter Summary
The Watchtower and the Map of Failures
From Monitoring to Observability: A Critical Distinction
The Taxonomy of Failure: A Field Guide to Drift
Building Your Proactive Warning System
In the Trenches and in the Boardroom
The Business of MLOps: Quantifying the ROI
Chapter Summary
Explainable AI (XAI)
Human-in-the-Loop (HITL) Frameworks: AI-Human Collaboration
Edge Computing and TinyML: Deploying Intelligence on Devices
Chapter Summary
The LLM Lifecycle: A New Set of Rules
LLMOps in Practice: Building the Guardrails
Monitoring LLMs in Production: A New Breed of Metrics
Field Story: Building an AI Co-pilot for Data Governance
An Architect of Intelligence
Chapter Summary
The FAANG Fallacy: Why You Can’t ‘Throw Money at the Problem’
The Architect’s Choice: Buy the Infrastructure, Build the Advantage
The Internal Data Platform: A “Data Ecosystem” for Your Enterprise
Navigating the Great Consolidation: Enclosure vs. Unification
Building and Scaling Data and AI Teams
The Economics of AI: Managing the FinOps Frontier
Chapter Summary
This practical guide is for data scientists, machine learning engineers, data leaders, and analytics professionals who want to move beyond notebooks, experiments, and one-time models. Analyze the real reason so many machine learning projects fail, and you will find that the problem is often not the algorithm. It is the data pipeline, the deployment process, the missing monitoring, the weak governance, or the lack of business ownership. This book shows how to treat models as living data products that must be designed, deployed, monitored, improved, and trusted.
Explore the full MLOps lifecycle, from data strategy and data contracts to model engineering, CI/CD pipelines, cloud infrastructure, model observability, and production machine learning. Design systems that can handle schema changes, data drift, feature drift, silent failures, unreliable data feeds, and changing business needs. Apply practical thinking to modern data platforms, data warehouses, data lakes, lakehouses, streaming architecture, automated retraining, model registries, and the tools that help data teams build dependable AI systems.
Evaluate the next frontier of applied AI with chapters on LLMOps, generative AI, prompt engineering, Retrieval-Augmented Generation (RAG), hallucination monitoring, explainable AI (XAI), Human-in-the-Loop (HITL) systems, and responsible AI governance. Create better enterprise AI applications by understanding how large language models change the deployment game while still requiring the same discipline, testing, observability, cost management, and accountability that define strong MLOps.
Assess your role not just as a model builder, but as an owner of business outcomes. The Deployed Data Scientist helps readers connect data science, machine learning, data governance, AI strategy, model deployment, cloud architecture, and business value into one practical roadmap. Whether you are building your first production model or leading a team responsible for enterprise AI, this book gives you the mindset, methods, and language to turn data science into systems that work.
Ankit Anand is a strategic Data Management Leader with over 17 years of experience designing and implementing enterprise-scale data solutions. He has led complex, global data initiatives at enterprises including Koch Industries and Lennox Industries, focusing on CI/CD automation, data governance, and modern data stack implementation. At Koch, he architected a pioneering multi-tenant, multi-domain MDM strategy and led the implementation of a robust CI/CD pipeline.
Dr. Scott Burk is the founder of It’s All Analytics. He is the author of nine books on AI, data science, and analytics, including the It’s All Analytics Series. He currently teaches in the MS in Data Science program at CUNY and has taught and developed curriculum at Baylor, Texas A&M, and SMU. He has solved complex AI, statistical, and analytical problems at Dell, Texas Instruments, PayPal, eBay, Overstock.com, healthcare organizations, energy firms, semiconductor and manufacturing companies, startups, and many others worldwide. Scott holds a BS in biology and chemistry, MS degrees in finance, statistics, and data mining, and a PhD in statistics. Data has been the unifying thread across his professional experience.
Kinshuk Dutta is a technology leader with nearly two decades of experience driving innovation in Data, Integration & Cyber Security. He is the co-author of the recent best selling Data for AI, founder of Be Cognizant of Data (datanizant.com) where he advocates for responsible data practices and AI enablement, a named inventor on AI patent. Kinshuk is a Senior Member of IEEE, recognized for shaping the future of intelligent systems and guiding organization through AI-driven transformation.
Please complete all fields.
