Boston AI Week is happening at the end of September, and I’m involved in three events. One roundtable I’m hosting, and two talks I’m giving. All three are free to attend, and all three are aimed at people who actually build software rather than people who just want to hear about it.
Here’s what’s on the schedule.
Architecting with AI: A Developer Roundtable on Software Engineering in the Agent Era
Saturday, September 26 · 10:00–11:30 AM ET Details and registration →
A roundtable on how autonomous multi-agent AI coding systems are changing the software engineering profession — architectural approaches, development workflows, evaluation methods, and what it means for engineers as AI takes on more of the writing.
I’m thrilled with the panel we put together for this one:
- CJ Avilla — Anthropic
- Chad Pytel — thoughtbot
- Kranthi Manchikanti — Microsoft
- Andrea Dulko — Google
- Jesse Waites — your host
Training Custom AI Models via Fine-Tuning
Tuesday, September 29 · 1:00–4:00 PM ET Details and registration →
This talk is a practical introduction to creating a custom AI model through fine-tuning. Rather than training a model from scratch, I will show how you can start with an existing open-weight model and specialize it to behave in a specific, repeatable way. The focus will be on the complete workflow: choosing a base model, defining the behavior you want to change, creating synthetic training data, establishing evaluation benchmarks, and fine-tuning with techniques such as LoRA and QLoRA.
I will also cover the parts of the process that are easy to overlook, including quantization, controlling variables during evaluation, checking for catastrophic forgetting, and measuring whether the fine-tuned model actually improved instead of simply assuming that it did. The goal is to demystify fine-tuning and show, step by step, how an open model can be turned into a specialized model that can be evaluated, deployed, and used in a real application.
The Multi-Agent Symphony: Operating Parallel Coding Agents with Conductor
Wednesday, September 30 · 6:00–8:00 PM ET Details and registration →
AI coding agents are quickly moving beyond the one-developer, one-agent model. Tools like Conductor, a Y Combinator-backed company building infrastructure for managing teams of coding agents, make it possible to run multiple agents simultaneously across isolated workspaces, assign them independent tasks, monitor their progress, and review and merge their work back into a real codebase.
In this practitioner-focused talk, we’ll explore what software development looks like when you begin managing a fleet of AI coding agents rather than working with them one at a time. Using Conductor as our primary example, we’ll cover parallel development with Claude Code, Codex, Cursor, and other agents, along with workspace isolation, Git worktrees, task decomposition, testing, code review, merge strategies, and the new coordination problems that emerge when several agents are writing code at once. We’ll also look beyond Conductor at other approaches to multi-agent development, focusing on the practical patterns that make parallel AI development manageable in real software projects.
We will also be giving out a limited amount of Conductor swag at the event!
Come say hi
If you’re in Boston that week, grab a seat at any of the three. Venue details are on each event page, and I’m always happy to talk shop before or after.
