React Under Fire: Rendering Performance for Streaming Agent UIs

In this session, I will cover the rendering and state management challenges specific to agentic UIs, drawing from production experience and from a public benchmark I built to measure rendering strategies under real agent event streams.

The session will address three areas. First, state architecture: how partial and final text differ as state shapes, why conventional Redux patterns struggle with high-frequency token streams, and what alternatives exist. Second, implementation tradeoffs: when a streaming library is justified, when a custom buffering layer is the better choice, I will demonstrate these comparisons through live coding, with performance measurements for each approach. Third, standards: what the AG-UI protocol currently specifies for agent-to-UI communication, and where it leaves decisions to the implementer.

Attendees will leave with a clear framework for choosing a rendering strategy, and access to the open benchmark to evaluate their own applications.

Beyond the Prompt: Designing Guardrails That Keep AI-Generated Code Maintainable

This session presents harness engineering: a structured approach to AI-assisted development that surrounds the model with guardrails rather than relying on better instructions. I will cover the workflow I use in a production codebase, a research and planning phase with a mandatory human checkpoint, an implementation phase, and a separate evaluation phase where the agent grading the work is not the agent that produced it.

The core of the talk is evaluation design: how to write grading criteria, why they should be weighted toward where the model is careless rather than where you care most, and how hard thresholds prevent one good score from masking a failure. I will demonstrate the full loop live against a real feature.
Attendees will leave with a workflow they can adapt, and a concrete framework for keeping AI-generated code to the standard their codebase already holds.

The Work Nobody Assigned You: Building Real Influence as an Engineer

We all very well know about the flaky test everyone learned to live with, the performance regression that sneaks up, so we file a bug that always stays in the backlog. The onboarding doc that would have saved the intern two weeks if someone had kept updating it. This is the work that sits in the gap between everyone's roadmap, and it's almost invisible, which is why it always stays undone - but high impact which is exactly why picking it up gets noticed.

But doing the work is only half of it, and it's the half most of us are comfortable with. The other half is the part engineers hate. You have to make it visible., and you have to make yourself visible - Not loud, not self-congratulatory, just visible enough that the people making decisions about your growth actually know it happened. A good piece of work that nobody knows you did moves your career exactly as much as work you never did.

In this talk I'll walk through what has actually moved the needle for me, across Microsoft and now Slack, and I'll say it from an engineer's seat, not a manager's. How to pick the off-roadmap work that builds trust instead of just eating your evenings. How to have the direct conversation about promotion and growth without it feeling gross. How to ask for real feedback, including from managers you don't report to, and then actually do something with it. And how the least glamorous thing on this list, a cold message to a team whose work you admire, tends to compound in ways nothing on your ticket board ever will.

Bonus: A personal agentic framework that silently keeps track of my wins - and does the documentation for my promo packet, a system that helps me win everyday at work

Target audience: Individual-contributor engineers, mid-level through senior, doing strong work who want it to translate into growth and influence. No prerequisites, valuable to anyone on an IC track.

Preferred session duration: 30 minutes

RIP, Chat Box: Designing AI Interfaces That Are Intuitive and Adaptive

It’s very difficult to talk to an LLM, and most of us don’t even realize we’re doing it wrong.

How good an answer is depends on how well you articulate the question. You’d have to explain your boundaries, share the data you have, ask it to research step by step, and bind its answers with a harness. Nobody does that every time. My last question to ChatGPT was “What are the best South-Asian scuba dive destinations for February?” and I got back a generic, text-heavy answer that I had to keep poking at.

That blank text box is the problem. We’ve handed users the hardest job in the product: figuring out what to ask.

In this session, I’ll walk through how to design AI interfaces that meet users halfway. We’ll break down the textbox anti-pattern, how to pick up intent from context the product already has, and how to swap long paragraphs for adaptive UI like options, filters and structured results that get users to what they need in fewer steps. Then we’ll get into A2UI, where agents respond with interface components, and what it means for frontend engineers building on top of it.