ZangesAiAssistant/AiAssistantBackend
This track equips learners to proficiently create, integrate, test, and deploy AI-driven applications and prompt engineering solutions, leveraging modern AI models and real-life cloud environments. This is a back-end oriented project.
Who this track is for?
- SE Students who want to gain experience and knowledge in AI and apply it for AI-Driven apps.
- Mid-High Proficiency
- Regular pace is OK
Example job postings
‣
Resources
Time to complete
- Between 2-3 months for MVP
Main Topics
- Fundamentals of AI, Machine Learning, LLM’s
- Evolution of AI - History of Machine Learning with examples of how it was used in each step.
- What changed with LLM’s - ChatGPT Revolution
- Current market leads and differences between them
- Prompt Engineering
- API Integration with AI models
- Structured Output w/ JSON
- Data Preprocessing and Cleaning
Guidelines for Choosing Idea
- The project must include data that is not part of the conversation and is dynamic (not in the system message).
- Prompt engineering project only - too simple.
- The project needs to have a clear business purpose and not just demonstrate abilities to use API.
- Create Automations that make processes more efficient.
Examples of projects
‣
‣
‣
‣
‣
Examples of bad projects
‣
‣
‣
Progress tracking
‣
‣
‣
‣
‣
1. Overview
The AI Assistant Dashboard is a web-based application designed to help users manage tasks such as note creation, calendar events, and document drafting through natural language interaction. The system leverages AI models for task processing and integrates with external services like Google Calendar.
2. Technology Stack
Backend
- Framework: FastAPI
- Database: PostgreSQL
- Calendar Integration & Authentication: Google API
- AI: OpenAI
Core Features
- Text Interface:
- A dashboard where users can type instructions to perform tasks like creating notes, scheduling events, and drafting documents.
- Task Routing:
- Implement a dispatcher that categorizes tasks (simple, intermediate, complex) and routes them to the appropriate AI model.(v2)
- Conversation History:
- Store user interactions in PostgreSQL with user ID, timestamps, and conversation context.
- Efficient Retrieval: (v2)
- Indexing: Create indexes on frequently queried fields (e.g., user ID, timestamps).
- Pagination: Limit results per query to ensure quick loading times.
- Caching: Utilize caching mechanisms for frequently accessed conversation data.
- Archiving: Plan for periodic archival of older conversation records to maintain performance.
https://dbdiagram.io/d/ai_assistant-67aa5fa3263d6cf9a0acfa58
https://docs.google.com/spreadsheets/d/1C3zDooiZcWw7Y2Qo1W4Hp7l_WuP18whwiygXKp9Mq8Q
Docs:
Google Calendar API overview | Google for Developers
unclecode/crawl4ai: 🚀🤖 Crawl4AI: Open-source LLM Friendly Web Crawler & Scraper
SQLAlchemy - The Database Toolkit for Python
Uncomplicated observability | Pydantic Logfire
Vertex AI Platform | Google Cloud