Architecture diagrams can explain what a system looks like, but they rarely explain how teams should evolve it safely. Python systems often grow through rapid iteration, integration needs, and team-specific decisions, and over time the challenge becomes less about drawing the right diagram and more about creating governance that helps people make better changes. This talk reframes software architecture as a living decision system, built around decision records, ownership boundaries, integration contracts, stakeholder alignment, and change readiness.
Code is cheaper than it's ever been, and it will never be this expensive again. Every few decades a technology flips the table on a profession, and software engineering is having that moment right now. The noise around it makes the moment hard to read, and most developers either haven't tried these tools at all or got slop and quit.
I spent the last year learning, building, and refining damn near everything with Claude Code, to unbelievable success. This is what survived contact with the work. The principles that held no matter which model was ahead that week, the tools I run on anything worth doing properly, and an honest accounting of which customizations earned their keep and which ones only ate my evenings. This presentation is for both of those people, plus anyone already using these tools who wants to compare notes.
During this presentation attendees will:
Learn where AI adoption lands in software engineering's long history of abstraction, and why this shift isn't the one the CASE tools promised
See a repeatable workflow that produces code they understand, and the guardrails that make an agent's output worth shipping
Leave knowing why cheap code makes them worth more, not less
In the past 10 years, I've mentored several really amazing young humans. I'm also a product of a lot of informal mentoring. Let me opine on what I've learned being on both sides of the desk and my general advice to anyone, regardless of your goals. We'll talk about:
How I got here
My general advice and how I set the stage in the first session
This talk is the story of how a small idea grew into a real, working social platform built around tiny 300-character posts. We built it mostly out of curiosity- using Python, a bit of GenAI, and a lot of trial and error. Along the way, we learned what actually matters when you're building something for real people, not just for yourself.
In October 2025, Python 3.14 was released and with it a cool new way to create t-strings, which promise to help python developers write better, and safer python code. While on the surface, they share a similar name to f-strings, the way we use them is quite different. This talk explores what they are, how they work, who they are for and how they will help python developers write safer code.
In an era where enterprises generate massive, complex streams of unstructured data from documents and logs to images, audio, and video, traditional search and analytics fall short. This session explores a transformative, production-ready framework for Multimodal Retrieval-Augmented Generation (RAG), designed to redefine how organizations unlock insights from every data modality. Discover how an intelligent agent, built using cutting-edge encoders and composable AI models, empowers organizations to understand, retrieve, and reason over diverse information in real time.
Merge Conflicts in software programming occur in 20% of all merges in open-source projects, incurring a median resolution time of ~6 minutes for a single conflict at best, or multiple days at worst. This talk will touch upon LLM Evaluations, Azure OpenAI Fine-Tuning, AI-generated resolution explanations, and usage of multiple Python SDKs to craft a scalable LLM-powered Merge Conflict Resolver. Our results indicate up to 125% performance improvement over vanilla LLM resolution accuracy.
Learn how to maintain code quality without sacrificing performance. This presentation explores SAST, SCAT, and SCA concepts, then introduces modern tools like Ruff and Pyright that make formatting, linting, and type checking simple and almost instantaneous.