ByteByteAi – 6 Weeks Cohort-based Course (Complete)

Source: https://bytebyteai.com/

The course is designed for those who want not only to study theory but also to build real artificial intelligence systems with their own hands. From language models to multimodal agents, you will follow the complete path of an AI engineer, creating working projects at each stage of learning.

What Awaits You

Project 1. LLM Playground

• Building your own sandbox for working with LLM
• Basics of language models: tokenization, architectures (GPT, Llama), text generation methods
• Post-training: SFT, RLHF
• Quality assessment methods: metrics, benchmarks, human evaluation

Project 2. Customer Support Chatbot on RAG and Prompt Engineering

• Model adaptation practice: fine-tuning, PEFT, LoRA
• Prompt engineering techniques: few-shot, zero-shot, chain-of-thought
• Retrieval-Augmented Generation: search, indexing, generation
• Evaluation of RAG systems

Project 3. “Ask-the-Web” Agent

• Building an agent that works with tools and the web
• Agent systems: planning, reflection, multiprocess workflows
• Tool calling and multi-agent approaches
• Methods for assessing agent efficiency

Project 4. Deep Research with Search and Reasoning Models

• Working with modern reasoning-LLM (e.g., OpenAI o1, DeepSeek-R1)
• Inference methods: CoT, Tree of Thoughts, self-consistency
• Training on reasoning data: SFT, RL with verifier, self-refinement

Project 5. Multimodal Agent (Text – Image/Video)

• Generation of images and videos: diffusion, GAN, VAE
• Architectures and training of diffusion models (U-Net, DiT)
• Quality assessment methods for generation: IS, FID, CLIP
• Building end-to-end T2I and T2V systems

Password: cms.ddpanda.org