Phi-3

AI Red Teaming Agent: Azure AI Foundry — Nagkumar Arkalgud & Keiji Kanazawa, Microsoft

Jun 27, 2025 · 19:31

Keiji Kanazawa and Nagkumar Arkalgud of Microsoft present the AI Red Teaming Agent in Azure AI Foundry, arguing that adversarial testing is essential for building trustworthy AI agents. Nagkumar demonstrates the tool: it runs scans against RAG apps or models directly, using attack strategies like Caesar encoding and Base64 to simulate adversarial prompts across four risk categories (violence, hate and fairness, etc.). In a demo with GPT-4.0 with full guardrails, no attacks succeeded; switching to Phi-3 without guardrails yielded 5 out of 40 successful attacks in hate and fairness. The tool integrates with Azure AI Foundry's content filters, which apply both input and output guardrails, allowing engineers to iteratively test and mitigate vulnerabilities. The talk emphasizes that red teaming should be part of a broader risk mapping and evaluation lifecycle, and that trust is a team sport requiring collaboration between engineers and security experts.

Multi model multimodal and multi agent innovations in Azure AI: Cedric Vidal

Feb 6, 2025 · 28:56

Cedric Vidal, Principal AI Advocate at Microsoft, demonstrates Azure AI's multi-model, multimodal, and multi-agent capabilities in a session packed with live demos. He shows how GPT-4 Omni mixes text and vision to read handwritten French menus and translate them, and diagnoses infrastructure damage from photos for energy and insurance industries. A new video translation service lets him speak German, Spanish, Italian, and Japanese in his own voice while preserving tone, such as whispering or yelling. The Azure AI model catalog now offers 1,600 models, including serverless deployment options, and Phi-3 Vision, a small 3.8B parameter model running locally in a browser via WebGPU. He also demonstrates code interpreter analyzing a GPX file from a kitesurfing session, plotting a map with turn markers, and GitHub Workspaces preview generating a Java GUI from Python code.

Creating and scaling your own custom copilots with Azure AI Studio: Hanchi Wang

Feb 6, 2025 · 24:21

Hanchi Wang, Software Engineer Lead at Azure AI, introduces Azure AI Studio and Promptflow for creating and scaling custom copilots, focusing on tracing, evaluation, and monitoring. He demonstrates a chatbot app that uses the Assistant API with a sales data insight tool (natural language to SQL) and a code interpreter. With Promptflow's trace decorator, developers capture inputs, outputs, and LLM interactions, viewable in a local UI and shareable via Azure AI Studio. For evaluation, he shows synthetic test data sets, content safety evaluators, and custom evaluators like execution time, error rate, and SQL similarity, comparing models such as GPT-4 Turbo, Mistral large, and Phi-3. In production, monitoring dashboards in Application Insights track model duration, token usage (prompt vs. completion), and failure rates, enabling engineers to optimize performance and cost.

Scaling AI in Education: A Khanmigo case study: Shawn Jansepar

Feb 5, 2025 · 22:39

Shawn Jansepar, Khan Academy's Director of Engineering, details building Khanmigo, an AI tutor and teacher assistant on GPT-4 with OpenAI, arguing generative AI can democratize one-on-one tutoring. He describes a rapid prototyping culture that launched Khanmigo in three months via a company-wide hackathon, replacing traditional agile with a prototype-to-beta-to-launch framework. Technical challenges include math accuracy through a math agent and chain-of-thought prompting, refactoring prompts into a component architecture for testing, and managing scale using multiple models and dedicated Azure compute. Jansepar highlights ethical design: a Socratic tutor that avoids answers, teacher moderation, and a writing coach with revision history. Khanmigo now has 200,000 paid users, half in school districts, with teacher tools sponsored by Microsoft free to US teachers. Plans include releasing a math tutoring benchmark and evaluating smaller models like Phi-3.

A Practical Guide to Efficient AI: Shelby Heinecke

Nov 18, 2024 · 17:45

Shelby Heinecke, who leads an AI research team at Salesforce, presents five orthogonal dimensions for making AI models efficient: efficient architecture selection, pre-training, fine-tuning, inference, and prompting. She highlights the power of small models like Phi-3 (3.8B parameters outperforming a 7B model), mobile LLM (350M parameters on par with 7B after fine-tuning), and Octopus (2B fine-tuned Gemma exceeding GPT-4 on Android tasks). For efficient inference, she explains post-training quantization, showing 4-bit quantization nearly halves memory usage without performance loss (e.g., LLaMA models), but warns 3-bit can degrade quality. She recommends frameworks like LLaMA CBP and ONNX Runtime for quantization and introduces her team's open-source Mobile AI Bench for evaluating quantized models, including an iOS app to measure latency and battery drain. The central claim is that deploying AI in constrained environments—cloud, on-prem, or edge—demands efficiency, and these practical techniques bridge the gap from demo to production.

No more bad outputs with structured generation: Remi Louf

Oct 14, 2024 · 15:32

Rémi Louf, CEO of .txt and co-maintainer of Outlines, argues that structured generation—guiding LLMs to produce valid regex, JSON, or context-free grammar outputs—eliminates parsing errors and hallucinations while adding negligible overhead. Outlines masks tokens that violate the target structure, enabling Mistral 7B to achieve 99.9% valid JSON (vs. 17% without). It also accelerates inference: a chatty 50-token ChatGPT answer shrinks to 8 tokens, and structured generation boosts open models beyond GPT-4—Phi-3 Medium hits 96.5% on Berkeley function calling (GPT-4 gets 93.5%). With one shot matching eight shots in accuracy, Louf contends that most text is structured and that structured generation should be the default for non-chatbot workflows.

Low Level Technicals of LLMs: Daniel Han

Jul 31, 2024 · 2:52:26

Daniel Han of Unsloth explains how to find and fix bugs in open-source LLMs like Gemma, Phi-3, and Llama, covering tokenizer issues, architecture pitfalls, and finetuning optimizations. He details the eight Gemma bugs Unsloth fixed, including a critical RoPE downcasting error that broke positional encoding, and a 2048 sliding window bug in Phi-3. Han walks through transformer internals: attention masking, layer norms, RoPE embeddings, and SwiGLU activation, showing how to derive gradients for custom kernels. He demonstrates Unsloth's 2x faster finetuning with 70% less memory via Triton kernels, and introduces new features: automatic Ollama model file creation, CSV fine-tuning with merged columns, and chunked cross-entropy for large vocabularies. The session includes live Q&A on learning rate schedules, precision trade-offs, and mechanistic interpretability.

Fixing bugs in Gemma, Llama, & Phi 3: Daniel Han

Jul 31, 2024 · 17:42

Daniel Han of Unsloth details eight bugs found in Llama 3, including double BOS tokens, untrained tokens in the base model, and pad tokens equaling EOS tokens, which cause infinite generations. He explains how Unsloth automatically fixes these issues and offers a free Colab notebook for fine-tuning with Ollama. Han also covers tokenization fixes for Gemma and a sliding window bug in Phi-3, emphasizing the importance of correct chat templates and avoiding double BOS tokens. The episode provides concrete code examples and best practices for fine-tuning open-source LLMs to avoid common pitfalls.