# Phi-3 Projects at AI Tinkerers

> Canonical HTML: https://aitinkerers.org/technologies/phi-3
> Markdown URL: https://aitinkerers.org/technologies/phi-3.md
> Technology record last updated: 2026-02-25T14:43:33Z
> Generated: 2026-09-22T11:45:42Z

Microsoft's family of small language models (SLMs) delivering high-reasoning performance on local devices and edge hardware.

Phi-3-mini packs 3.8 billion parameters into a footprint small enough for local deployment on an iPhone 14. Trained on a 3.3 trillion token dataset of high-quality synthetic data and filtered web content: it outperforms models twice its size (like Mixtral 8x7B) on benchmarks for coding and logic. The family includes 7B (small) and 14B (medium) variants, providing developers with low-latency options for complex tasks without the massive compute requirements of traditional LLMs.

- Official technology site: https://azure.microsoft.com/en-us/blog/introducing-phi-3-redefining-what-is-possible-with-slms/
- Public AI Tinkerers demos and talks: 2
- Result page: 1 of 1

## Recent Public Talks and Demos

### [Culturally Aligned AI: Building Dlab-852-Mini for Hong Kong Cultural Nuances](https://hong-kong.aitinkerers.org/talks/rsvp_nJ6fQuisHzw)

I'll showcase our work at Decisions Lab on Dlab-852-Mini, a specialized fine-tuned version of Microsoft's Phi-3.5-mini-instruct model designed to accurately replicate and align with Hong Kong's unique cultural perspectives, attitudes, and behaviors. Drawing from our cultural alignment research, we'll explore how we created the CultureKit eval, a suite of CLI for assessing cultural biases in LLMs, to train and evaluate the model. I'll cover the fine-tuning process, benchmark results demonstrating up to 2x better performance in simulating local responses compared to base models, and practical case studies. Project URL - Website or Github (Optional) https://github.com/decisionslab/culturekit Project URL 2 (Optional) Video Demo URL (Optional) https://www.youtube.com/watch?v=example-demo (replace with actual if available; otherwise, omit)

- Event context: AI Tinkerers Hong Kong: Deep Dive Series — Technical Lessons from Shipping AI/ML/CV — 2025-12-18 — Hong Kong
- Public talk page: https://hong-kong.aitinkerers.org/talks/rsvp_nJ6fQuisHzw

### [AutoML Agent: Automating ML Code Generation](https://toronto.aitinkerers.org/talks/rsvp_D5m-tfkILPA)

This demo will showcase a custom proof of concept AI Agent that autonomously generates and executes Python code to find the best machine learning model for a given problem. Built from scratch this agent addresses limitations common in agentic frameworks like Langchain or CrewAI, providing enhanced flexibility and control for production-grade applications. 1. Live Demo We'll begin with a live Colab notebook demo, which attendees can access to try the application during the presentation or later. Built entirely in Python with open-source libraries and LLMs, this demo enables hands-on participation. 2. Key Components We’ll then walk through the application’s architecture, covering: State Management: Using the transitions library, we manage the state machine workflow to handle tasks such as entity extraction, code generation, error handling, and re-execution, ensuring modular and resilient task transitions. Open-Source LLM: We use an open source Phi 3.5 model for core tasks like code generation, entity extraction, and error management, demonstrating the capabilities of open-source LLMs for AutoML. 3. Production Architecture Overview Finally, a production-ready architecture, visualized on Miro, demonstrates an end-to-end production solution on AWS, including: Frontend and API Layer: Allowing user interaction through a secure, scalable API and frontend interface. Load Balancing and Model Deployment: AWS Lambda with load balancing optimize for scalable, responsive model deployment. Security: VPCs, and user verification via AWS Cognito ensure controlled access and data security. Logging and Monitoring: Model monitoring and tracing via MLFLow on AWS Sagemaker.

- Event context: AI Tinkerers Toronto - November 2024 Meetup at Accenture — 2024-11-28 — Toronto
- Public talk page: https://toronto.aitinkerers.org/talks/rsvp_D5m-tfkILPA

## Related Technologies

- [AWS Lambda](https://aitinkerers.org/technologies/aws-lambda) ([Markdown](https://aitinkerers.org/technologies/aws-lambda.md)) — 14 public demos
- [AWS SageMaker](https://aitinkerers.org/technologies/aws-sagemaker) ([Markdown](https://aitinkerers.org/technologies/aws-sagemaker.md)) — 2 public demos
- [datasets](https://aitinkerers.org/technologies/datasets) ([Markdown](https://aitinkerers.org/technologies/datasets.md)) — 6 public demos
- [LangChain](https://aitinkerers.org/technologies/langchain) ([Markdown](https://aitinkerers.org/technologies/langchain.md)) — 445 public demos
- [MLflow](https://aitinkerers.org/technologies/mlflow) ([Markdown](https://aitinkerers.org/technologies/mlflow.md)) — 5 public demos
- [MLX](https://aitinkerers.org/technologies/mlx) ([Markdown](https://aitinkerers.org/technologies/mlx.md)) — 7 public demos
- [Python](https://aitinkerers.org/technologies/python) ([Markdown](https://aitinkerers.org/technologies/python.md)) — 662 public demos