AI Residency

This repository contains notebooks, scripts and other resources from the DDS AI Residency program. Materials range from Python fundamentals and deep learning tutorials to examples using frameworks such as LangChain and LangGraph. Use these samples and demos as references for building chatbots, exploring retrieval‑augmented generation (RAG) techniques and experimenting with agentic workflows.

DDS AI Residency — Code, Notebooks & Demos

This repository is the hands-on lab archive for the Decoding Data Science (DDS) AI Residency program.
It contains the exact notebooks, scripts, and mini-projects used across cohorts — from Python foundations and prompt engineering, through RAG + evaluation, and ending with agentic systems + MCP.

Recommended flow: Prompting → Python + Data → LLM Apps → RAG → Evaluation → Agents → Multi-Agents → MCP → Ship

---

Quick Start

Open in Google Colab (no installation)

1. Go to: https://colab.research.google.com
2. Open any notebook directly from GitHub:
   • Replace the URL pattern:
     https://colab.research.google.com/github/Decoding-Data-Science/airesidency/blob/main/<NOTEBOOK>.ipynb

Repo Navigation (Folders)

• python_for_ai/ — Python fundamentals for AI builders
• fast_api/ — API deployment demos
• aws-bedrock/ — Bedrock examples + integrations
• langchain_airesidency/ — LangChain apps, tools, memory, agents
• langgraph/ — LangGraph workflows, corrective RAG, agent graphs
• ragmc/ — Retrieval-Augmented Generation demos and variants
• mcp_handson/ — MCP hands-on labs (late-stage: tool ecosystem + integration)
• crewai-first/ — Multi-agent systems with CrewAI
• kahoot/ — Quizzes used in sessions
• math_llm/ — (Historical) LLM + math experiments

---

Program Map (Day-by-Day Learning Path)

This is the suggested day-level flow we use in the Residency.
Each day points to the type of files you’ll find in this repo and how they connect.

Day 0 — Builder Setup + Prompt Engineering Foundations

Goal: stop “prompt luck” and build repeatable prompting habits.

What you learn

• Zero-shot / few-shot / structured prompting
• System vs user instructions, grounding, constraints
• Output formatting, safety & guardrails mindset

Where to look in this repo

• tools_agents.ipynb, tools_multi_agents.ipynb (tools + thinking patterns)
• Notebooks that contain “assistant” patterns (e.g., weather assistants, first chatbots)

---

Day 1 — Python for AI Builders

Goal: write clean Python to ship AI apps fast.

Key notebooks

• Variables.ipynb
• Lesson_2_Repeating_tasks_with_for_loops.ipynb
• Using_files_in_Python.ipynb
• strating_with_data_in_python.ipynb
• xtracting_restaurant_information_from_journal_entries.ipynb
• Lesson_3_Reading_journals_from_food_critics.ipynb

Outcome

• You can manipulate text/data, structure outputs, read files, and prepare inputs for LLM pipelines.

---

Day 2 — From Python to Your First LLM App (Chatbots)

Goal: build a working chatbot end-to-end, then ship a UI version.

Core demos

• first_chatbot_19jul.ipynb
• first_chatbot_19jul_gradio.ipynb
• first_python_chatbot_logo_HF.ipynb
• python_tutor_03_01.ipynb, python_tutor_22_02.ipynb (teaching/assistant style bots)

LLM providers used across cohorts

• Groq demos:
  • groq_colab.ipynb, groq_colab_5july.ipynb, groq_sdk.ipynb
  • groq_colab_5july_withgradio.ipynb
• Bedrock demos:
  • aws-bedrock/

Outcome

• You can move from notebook → working assistant → basic Gradio/HF deployment.

---

Day 3 — Embeddings & Vector Databases

Goal: understand embeddings as the foundation for retrieval, memory, and agent grounding.

Key notebooks

• Vector_embedding.ipynb
• weavite_Vector_embedding23_jan.ipynb
• weavite_Vector_embedding29mar.ipynb

Outcome

• You can generate embeddings, store/search vectors, and prepare for RAG.

---

Day 4 — RAG: Your First Retrieval-Augmented Generation System

Goal: stop hallucinations by grounding responses in your documents.

Core notebooks

• your_firstRAG.ipynb
• your_firstRAG_cohort7.ipynb
• your_firstRAG_cohort7_updated.ipynb
• rag_llamaindex.ipynb
• Copy_of_rag_llamaindex.ipynb
• first_rag_llamaindex_c7.ipynb

Index + vector store demos

• Cohort_6_PineconeIndexDemo.ipynb
• Copy_of_PineconeIndexDemo.ipynb
• MC_11_PineconeIndexDemo.ipynb

Outcome

• You can ingest docs → chunk → embed → retrieve → answer with citations/grounding.

---

Day 5 — LLM Evaluation & Quality

Goal: measure quality like an engineer, not by vibes.

Key notebooks

• 1_evaluation_recipe.ipynb
• 1_evaluation_recipe-c7.ipynb

Outcome

• You can define evaluation signals (accuracy, relevance, faithfulness), test changes, and iterate systematically.

---

Day 6 — LangChain Core (Chains, Tools, Memory)

Goal: turn prompts into applications with structure: tools, memory, retrieval, routing.

Key notebooks

• langchain_agent.ipynb
• Memory_management_for_AI_Apps_Agents_using_Langchain.ipynb
• Memory_management_for_AI_Apps_Agents_using_Langchain6thsep.ipynb
• LangChain chain + agent variants:
  • Langchain_chains_agent_6thsep.ipynb
  • Langchain_chains_agent_12thapr.ipynb
  • Langchain_chains_agent_12thjul.ipynb
  • Langchain_chains_agent_15thov.ipynb
  • Langchain_chains_agent_1feb-2025.ipynb
  • Langchain_chains_agent_1febnew.ipynb

Outcome

• You can build tool-using assistants and control behavior with memory and routing.

---

Day 7 — LangGraph: Agentic Workflows & Corrective RAG

Goal: build stateful agent systems with robust control flow.

Key notebooks

• simplegraph20sep.ipynb
• simplegraph20thfeb.ipynb
• Copy_of_simplegraph20sep.ipynb
• Langraph_with_tools_cohort7.ipynb
• Building_an_Agentic_Corrective_RAG_System_with_LangGraph.ipynb

Outcome

• You can model multi-step workflows (states, retries, tool calls, verification loops).

---

Day 8 — Multi-Agent Systems (CrewAI)

Goal: orchestrate specialists that collaborate on a goal (planner/writer/reviewer/tool-runner).

Key notebooks

• crewai_21stnov2024.ipynb
• crewai_15thfeb.ipynb
• crewai_26th.ipynb
• crewai_multiagent_2025_level_1.ipynb
• crewai_multiagent_2025.ipynb
• crewai_multiagent_Gitex_level_2_.ipynb
• crewai_multiagent_Gitex_level_2_29th_nov_.ipynb
• multi_agent_collaboration.ipynb

Reference PDF

• Ship_Your_Multi-Agent_System_Building_a_Collaborative_Writer_App_with_CrewAI.pdf

Outcome

• You can design roles, tasks, tool access, and evaluation for multi-agent pipelines.

---

Day 9 — MCP Hands-On (Late Stage)

Goal: connect agents to real tools reliably, like production systems.

Where to look

• mcp_handson/

Why MCP comes late

• MCP becomes powerful only after you understand:
  • tool calling + structured outputs
  • routing + memory
  • evaluation + safety boundaries
  • multi-agent orchestration

Outcome

• You can standardize tool integrations and scale beyond one-off scripts.

---

Applied Demo Tracks (Real Use Cases)

These notebooks are used as mini-projects and concept reinforcement.

Weather Assistants (tools + function calling + deployment readiness)

• Weather_assistant26jul.ipynb
• weather_assistant14thsep.ipynb
• weather_assistant24thmay.ipynb
• weather_assistant4thoct.ipynb
• Weather_functioncall (1).ipynb
• Another_copy_of_weather_assistant.ipynb

Financial Market Assistants

• stock_market_14thsep.ipynb
• stock_market_26_jul.ipynb
• stock_market_28th_sep.ipynb
• Folder: Creating AI Financial Market Assistant Using AI Agents/

Fine-tuning / Transformers (Optional / Advanced)

• hf_transformer.ipynb
• huggingface_practice.ipynb
• Text_Summary_T5_Fine_Tuned.ipynb
• lora_finetuning.yml

---

Recommended Learning Order (If You’re Self-Studying)

1. Prompting & tool mindset
2. Python foundations (files, loops, data basics)
3. First chatbot + Gradio (ship something)
4. Embeddings
5. RAG + Pinecone/LlamaIndex
6. Evaluation recipes
7. LangChain (tools/memory/routing)
8. LangGraph (workflows + corrective loops)
9. CrewAI multi-agents
10. MCP hands-on (production-grade tool ecosystem)

---

How to Use This Repo During Cohort

• Use notebooks as reference implementations (not copy-paste only).
• Every major concept has at least one of:
  • a minimal demo
  • a cohort-specific variant
  • a shipping version (Gradio / deployment)

Tip for trainers: Prefer the most recent cohort notebook variant when duplicates exist.

---

License / Usage

These materials are provided for DDS AI Residency learning and internal cohort use.
If you want to reuse for training delivery or workshops, request permission from DDS.

---

Support

• Ask in the cohort community channel (preferred)
• Tag your mentor with the notebook name + error screenshot + what you tried