Learning Agentic AI: A Personal Journey of AI and Machine Learning

Project Basic Information

• Original Author/Maintainer: mateusmtorres23
• Source Platform: GitHub
• Original Title: Learning-Agentic-AI
• Original Link: https://github.com/mateusmtorres23/Learning-Agentic-AI
• Release Time: 2026-05-29

Core Introduction This project documents developer mateusmtorres23's personal learning journey in artificial intelligence and machine learning, serving as a reference path for beginners to enter AI. With "Agentic AI (Autonomous Agent AI)" as its core theme, the project practices the concept of "Learning in Public", covering key technical directions such as LangChain and LLM, and provides learners with an exploration framework from basics to advanced levels.

In the field of technical learning, "Learning in Public" has become a popular concept. By documenting and sharing the learning process, learners can consolidate knowledge, provide references for others, build personal brands, and integrate into technical communities.

mateusmtorres23's Learning-Agentic-AI project is an embodiment of this concept; it is a personal learning repository that records the author's exploration journey in artificial intelligence and machine learning.

Definition

Agentic AI (Autonomous Agent AI) refers to AI systems that can independently perceive the environment, make decisions, and execute actions, contrasting with traditional passive AI (such as classifiers and recommendation systems).

Core Features

• Goal-oriented: Plans and executes around specific goals
• Environment interaction: Bidirectional interaction with the environment
• Continuous operation: Runs long-term instead of single response
• Adaptability: Adjusts behavior based on feedback
• Tool usage: Calls external tools and APIs

Differences from Traditional AI

Traditional AI	Agentic AI
Single inference	Continuous decision loop
Passive response	Active action
Fixed input/output	Dynamic environment interaction
Single task	Multi-step task chain

Agentic AI is a hot direction in the current AI field; the popularity of projects like AutoGPT and LangChain is a reflection of this trend.

Based on the theme of "Agentic AI", the project's learning path may cover the following stages:

1. Foundation Stage

• Python programming: Main language for AI development
• Math basics: Linear algebra, probability theory, calculus
• Machine learning basics: Supervised learning, unsupervised learning, model evaluation

2. Deep Learning

• Neural networks: Feedforward networks, backpropagation
• Deep learning frameworks: PyTorch or TensorFlow
• Computer vision/NLP: Choose direction based on interest

3. Large Language Models (LLM)

• Transformer architecture: Understand attention mechanism
• Pre-training and fine-tuning: BERT, GPT series models
• Prompt engineering: Techniques for effective LLM use

4. Agent Development

• ReAct mode: Combining Reasoning + Acting
• Tool usage: Enable models to call external APIs
• Memory mechanism: Short-term and long-term memory design
• Multi-agent systems: Collaboration among multiple AI agents

5. Frameworks and Tools

• LangChain: Framework for building LLM applications
• AutoGPT: Implementation of autonomous AI agents
• Vector databases: Store and retrieve embedded vectors

For Learners Themselves

• Knowledge consolidation: Teaching is the best way to learn; writing forces deep thinking
• Progress tracking: Visualize learning journey, enhance sense of achievement
• Community feedback: Get suggestions and guidance from others
• Job-seeking advantage: Demonstrate learning ability and technical enthusiasm

For the Community

• Learning resources: Provide reference paths for later learners
• Problem discovery: Record pitfalls encountered to help others avoid them
• Technical exchange: Trigger discussions and promote knowledge dissemination

If you want to start your "Learning in Public" journey, you can refer to the following suggestions:

1. Define Goals

• Determine learning theme (e.g., Agentic AI, MLOps)
• Set time frame (e.g., 6 months, 1 year)
• Define measurable milestones

2. Content Organization

• README: Project overview and learning roadmap
• Notes: Markdown files organized by topic
• Code: Source code of practice projects
• Resources: Collected useful links and materials

3. Continuous Updates

• Submit learning progress regularly
• Record encountered problems and solutions
• Share learning insights and reflections

4. Community Interaction

• Share updates on social media
• Participate in relevant community discussions
• Respond to Issues and Pull Requests

Papers and Theory

• ReAct: Synergizing Reasoning and Acting in Language Models
• Toolformer: Language Models Can Teach Themselves to Use Tools
• AutoGPT-related papers

Open Source Projects

• LangChain: Standard framework for building LLM applications
• AutoGPT: Early representative of autonomous AI agents
• BabyAGI: Task-driven autonomous agent
• MetaGPT: Multi-agent collaboration framework

Online Courses

• DeepLearning.AI's LangChain course
• Hugging Face's Transformers course
• Various YouTube tutorials and blog posts

Summary

Although mateusmtorres23's Learning-Agentic-AI project is concise, it represents a positive attitude towards technical learning. In today's rapidly evolving AI technology landscape, continuous learning and public sharing are particularly important. Whether you are new to AI or not, building and maintaining a personal learning repository is a recommended practice—it not only records your growth trajectory but also contributes value to the community, achieving a win-win situation.

Outlook on Technical Trends

Agentic AI represents an important direction in the development of artificial intelligence. With the improvement of LLM capabilities, we can expect:

• Stronger reasoning abilities: Complex planning and decision-making
• Richer tool ecosystems: More tools that agents can call
• Better memory mechanisms: Improved long-term memory and knowledge management
• Multimodal agents: Processing text, images, audio, etc.
• Agent collaboration: Multiple agents working together to complete complex tasks