Tech Events & Workshops

A practical two-hour session on the principles behind document question-answering systems, with Rust as the implementation context. The session focuses on how applications can answer questions from a known set of documents instead of relying only on what a language model already knows. We will cover the core workflow: loading documents, preparing them for search, finding relevant passages, passing useful context to a model, and producing answers that stay connected to the source material. The goal is to teach the engineering ideas behind RAG systems without turning the session into a product demo or an abstract AI lecture. The emphasis is on system boundaries, retrieval quality, answer grounding, error handling, and the practical decisions developers need to make when building document-aware applications. Audience: entry-level and intermediate developers who want a practical session on document search, retrieval, grounded generation, and Rust application design. Outcomes: * Understand how document-based question answering systems work * Explain the purpose of chunking, embeddings, retrieval, and context assembly * Design a simple retrieve-first, answer-second workflow * Use Rust to organize the main parts of a small RAG application * Recognize common causes of poor or unsupported answers * Apply basic techniques for improving retrieval and answer quality * Think about how to evaluate whether the system is actually useful Format: two hours with a short framing walkthrough, a concrete example, discussion of tradeoffs, and a closing checklist for practice.

We've all seen it: you ask an AI for a simple feature, and it returns a 50-line wall of boilerplate filled with nested interfaces and redundant logic. As engineers, we still have to read, understand, and debug that code. The more "noise" in our stack, the more room there is for hallucinations, and the harder it is for us to spot them. This Saturday, we're exploring how to cut through that noise. We'll look at Clojure(Script), a language where code is data. By reducing the surface area of our syntax, we don't just save tokens; we gain a level of clarity that makes it easier for humans to reason about their systems and for AI to stay on track. *The "Honesty" Stack:* - **Malli as the Source of Truth:** Learn how shared schemas act as a "Contract of Honesty." Whether a human wrote the code or an LLM generated it, Malli ensures the data remains valid before it ever touches your logic. - **Hiccup & Replicant:** See how UI development changes when your DOM is just a data structure. We'll use Replicant to demonstrate how pure-data templates (Hiccup) lead to high-performance, predictable interfaces without the "magic" of complex frameworks. - **HoneySQL:** Why write strings when you can write maps? We'll show how database queries become searchable, composable data, keeping your persistence layer as clear as your business logic. - **The Living REPL:** Experience a tight feedback loop where you evaluate snippets instantly in a running environment. It's the ultimate tool for verifying behavior and ensuring that what you (or your AI) just wrote actually works. ## Why Attend? If you're looking for a way to build robust, verifiable systems without the weight of traditional boilerplate, this talk is for you. We aren't here to talk about AI doing your job - we're here to talk about how Clojure(Script) makes you a more effective architect of the code you produce.