Dandekar-HI1107×1388 225 KB

Raj Abhijit Dandekar, Rajat Dandekar, Sreedath Panat, Naman Dwivedi

Build a DeepSeek Model (From Scratch) is a hands-on guide to creating your own DeepSeek-style model step by step. You’ll start with a base LLM, then implement reasoning, retrieval, and optimization components to reproduce DeepSeek’s key architectural ideas. It’s an accessible deep dive into how multi-stage inference, reasoning traces, and reinforcement loops come together to create models capable of logical, multi-step problem solving.

In this book, you’ll learn how to:

• Implement a DeepSeek-style reasoning framework from the ground up

• Integrate retrieval augmentation, reasoning tokens, and self-reflection

• Apply reinforcement learning for reasoning improvement

• Measure reasoning performance across math, logic, and code tasks

• Understand how reasoning models differ from traditional LLMs

Whether you’re a machine learning engineer, researcher, or developer curious about reasoning-centric AI, Build a DeepSeek Model (From Scratch) offers a transparent, implementation-first look into one of the most exciting frontiers in modern AI.

• Full details: https://www.manning.com/books/build-a-deepseek-model-from-scratch

Don’t forget you can get 45% off with your Devtalk discount! Just use the coupon code “devtalk.com” at checkout