| |
|
Daytona & Sambanova AI Builders
|
| With Varshika Gambhir (Research Enggr, Google Labs), Vashishtha Patil (Applied Scientist, Amazon), Muhammad Annas Hashmi (DevRel, Daytona), Vasanth Mohan (Head of Dev Rel & Product Mktg, SambaNova), Sako M (S/w Enggr, You.com). |
|
Sep 23 (Wed) @ 05:30 PM
FREE
| | Venue, To Be Announced, SF
|
|
|
|
|
| |
| Register |
|
|
| |
An event dedicated to exploring all things AI Engineering!
Event partners: SambaNova, You.com & WeAreDevelopers
Agenda
5:30 pm - 5:35 pm
Welcome & Opening Remarks
Marijan Cipcic, Principal Events Manager at Daytona
5:35 pm - 5:50 pm
Talk "Your Agent's Slowest Tool Is Its Mouse"
Muhammad Annas Hashmi, DevRel at Daytona
Outline:
Every click a GUI agent makes costs a full screenshot & a model round trip, & it only lands if the layout stayed where the model last saw it. Chain a task out of clicks & you have bought seconds of waiting & a context window full of pixels for work a shell one-liner could have done. Screenshots are heavy. Text is light. And a click sequence is the most fragile program ever written. No variables, no error handling, & the only retry is asking the model again.
This talk builds the cost model of computer use. Where the waiting actually comes from, why pixel agents break when nothing is wrong, & the levers that fix it: read structure (a DOM, an accessibility tree) instead of rendering it to an image first, write code instead of emitting one click at a time, & cache what worked so the second run costs nothing. I'll then follow up with a demo of what we're doing at Daytona to address these.
5:50 pm - 6:05 pm
Talk "Fast Tokens, More Responsive Agents"
Vasanth Mohan, Head of Dev Rel & Product Marketing @ SambaNova
Outline:
Waiting on an AI agent kills the magic. Every time it reasons, generates, calls a tool, checks the result, & goes again, with today's infrastructure, latency stacks up. This lighting talk takes a look at how matching the right processor to each part of the job-GPUs for prefill, Reconfigurable Dataflow Units (RDUs) for fast decoding, & CPUs for orchestration & tool calls-can keep those loops moving. The goal is simple: agents that feel fast, fluid, & actually fun & productive to use.
6:05 pm - 6:20 pm
Talk "TBA"
Sako M, Staff Software Engineer at You.com
Outline:
TBA
6:20 pm - 6:30 pm
Talk "Beyond the Sandbox: Scaling Enterprise-Grade AI Agents for Real Users"
Varshika Gambhir, Staff Research Engineer at Google Labs
Outline:
We've all seen the demos: a large language model performing magical feats in a perfectly controlled environment. But what happens when that AI agent leaves the sandbox & meets the messy reality of enterprise workflows, strict compliance, & unpredictable customers?
Today, the industry is stuck in the "prototype graveyard." Moving from a slick proof-of-concept to a reliable, production-ready agentic product requires a fundamental shift in how we build & evaluate software.
Drawing on my 01 R&D experience at Google Labs-including architecting Google's first cross-modality AI agent, Ask Advisor-this keynote bridges the gap between foundational ML research & massively scalable infrastructure. We will explore the technical & strategic playbook required to transform legacy systems into robust, multi-billion-dollar agentic solutions.
Key Takeaways:
Architecting for Agency: Designing cross-modality, multi-agent architectures that survive complex enterprise environments.
Evaluating the Unpredictable: Implementing advanced LLM-as-a-judge evaluation loops to guarantee reliability.
The Deterministic Bridge: Wrapping non-deterministic AI in strict guardrails, state management, & fallback mechanisms.
01->100 to Massive Scale: The operational blueprint for moving R&D prototypes to trusted, customer-facing products
6:30 pm - 6:40 pm
Talk "Auto-Research on a Budget: Small Models in the Loop, Frontier Models on Call"
Vashishtha Patil, Senior Applied Scientist at Amazon
Outline:
Autonomous research agents that propose, implement, & refine ML solutions have gotten remarkably good. They have also inherited an assumption that excludes most teams: a frontier model drives every step of the loop. That assumption matters because loop cost, not model quality, is becoming the limit on how much autonomy a team can afford to run.
This talk explores inverting it. A small open-weight model runs the loop, & a frontier model is called in only as an advisor, on a metered budget.
We'll cover what the literature establishes about small & large model collaboration, including step-level escalation, agent distillation, & budget allocation, & where it stops short for long-horizon loops, whose failure modes are not bad tool calls but dead branches, validation leaks, & unclear stopping points.
We'll look at a framework for deciding when advice is worth buying, how much of a trajectory an advisor needs to see, & how to measure guidance rather than assume its value. We will walk through a recorded run showing the loop, the trigger, & the cost meter together.
The session concludes with open problems, including asynchronous advisors, persistent guidance, & what cost-aware autonomy means for teams without frontier budgets.
6:40 pm - 8:30 pm
Networking
With pizzas & beverages
About event
This is dynamic gathering for AI enthusiasts, innovators, & professionals to collaborate, share ideas, & explore the latest advancements in artificial intelligence. Whether you're building AI products, researching cutting-edge algorithms, or simply passionate about the field, join us to connect, learn, & drive the future of AI forward.
|
|
|
|
|
|
|
|
