AI Is Not Magic: From Prompts to Your First Fine-Tuned Classifier
Saturday, May 9
3:00 PM, UTC1 hour
LiveFree to Join
Sign in or create an account to watch the recording
Register once to get unlimited, free access to the free sessions
What you'll learn
The Evolution of Classification: Tracing the journey from brittle Regex and IF/THEN conditions to Prompting and RAG—and why they cap out at 80% accuracy.
The Fine-Tuning Fix: How custom models solve the consistency, format control, and scaling issues that prompts leave behind.
The Decision Framework: How to weigh cost, latency, and accuracy to know exactly when to make the jump to a custom model.
Data & Training: Prepping real-world logs to train your first classifier while avoiding bad labels and overfitting.
Prototype to Production: Evaluating, debugging, and safely deploying your AI classifier like a software engineer.
Why you should attend
Be honest: how much of your week is spent begging an LLM API to please just return a valid JSON object? For most software engineers, calling an AI API feels safe, while "fine-tuning" sounds like dark magic reserved for ML researchers. But when standard prompts cap out at 80% accuracy, and unpredictable text breaks your downstream parsers, you need a better way.
We are cutting through the hype to show you that if you know how to write a unit test, log an error, or deploy a microservice, you already have the skills to build a custom classifier. Join us to stop relying on "vibes-based" prompt engineering and learn how to apply true software engineering rigor to build reliable, production-ready AI.
Session requirements
Technical Baseline: Comfort with standard REST API requests, parsing JSON, and any programming language (code examples will be in Python, but the concepts apply universally).
Prior AI Experience: You do not need a machine learning background or a math PhD. You just need to have tried—and been frustrated by—standard prompt engineering.
What to Bring: Bring your laptop if you want to follow along with the code examples, and a readiness to treat AI as a standard software component rather than a black box.
More about hosts & guests
I am a Staff Software Engineer with twelve years of full-stack development expertise. Throughout my career, I have driven high-impact automation and scale, including leading the technical implementation of a multi-billion dollar enterprise project, as well as automating a health insurance platform using AI that increased operational efficiency by 300%.
I build top-ranking open-source AI tools. I am the creator of Chatterbox, an Egyptian Arabic TTS model, and an English-to-Egyptian Arabic translation model that ranked second just behind GPT-4o on the national leaderboard. Currently, my active focus is on LAHGTNA, an ongoing project to build a unified speech synthesis model that comprehensively covers all Arabic dialects.
Previously, I was a Co-Founder of the open-source NAMAA Community, where I led the development of multilingual AI tools like the QARI Arabic OCR model. I am also a published researcher with papers on Multimodal Large Language Model.
160 students registered 🔥Join them now