CALLING ALL PIONEERS:
JOIN OUR STAGE

September 29 – October 1, 2026
Moscone South · San Francisco, CA

About the conference

Where AI leaders share what actually works

Fully Connected 2026: Built for AI is CoreWeave's global flagship conference and the destination for the engineers, researchers, and platform leaders running AI in production. Every session is practitioner-first: real deployments, real infrastructure decisions, real outcomes. If you've built something worth sharing, this is your room.

2,000+

Live Attendees

Technical Tracks

30+

Breakout Sessions

300+

Minimum Session Level

Tracks

3 tracks. 30+ sessions. Find your stage.

Track 1

Build adaptive AI infrastructure

The compute, networking, storage, and platform tooling powering AI at scale. Hear real-world lessons from teams running production AI infrastructure and data platforms.

Sessions
Available

Track 2

Push models further

From training to inference, explore how leading AI teams build, optimize, and deploy modern AI workloads in production using CoreWeave and W&B tooling.

Sessions
Available

Track 3

Evaluate and monitor agents

Explore how teams are building, evaluating, and improving autonomous AI agents in production. Sessions cover agent development, observability, RL workflows, and self-improving systems.

Sessions
Available

Why Speak

Built something worth sharing with the world?

This is where the most credible AI practitioners in the world come to share what's actually working. A session at Fully Connected is not a marketing slot—it's a clear signal to your fellow pioneers that you've built something worth the room's attention.

The right audience for your work

2,000+ engineers, researchers, and platform leaders who evaluate ideas based on technical merit, not company size or title.

Reach beyond the room

Sessions are professionally recorded and distributed post-event—extending your reach to tens of thousands of practitioners who weren't in San Francisco.

Peer credibility that lasts

Speaking at Fully Connected signals that your work is production-grade and peer-reviewed. It's the credential that opens doors in the AI community long after the event ends.

Access to the CoreWeave + Weights & Biases ecosystem

Connect directly with the CoreWeave and W&B teams—engineers, product leads, and researchers building the platform your work depends on.

Complimentary full conference pass

Selected speakers receive a complimentary General conference pass — full access to all sessions, keynotes, hands-on labs, evening receptions, and the headliner concert.

Session formats

The right format doesn't just fit your story, it accelerates it

Four ways in, one requirement: show your work

Technical deep dive + demo

45 mins
Presentation (30) + Q&A (15)

One topic, explored at depth. Code, architecture diagrams, benchmarks, and the decisions behind them. The format for practitioners who want to go beyond the headline and into the implementation. A live demo is required.

Live demo + deep dive

45 mins
Demo (20) + Q&A (25)

Show something working in a live environment, with real data. No slides required. The format for builders who want to prove the thing, not describe it. The demo runs first; the discussion follows.

Panel discussion

45 mins
Moderated

3–4 practitioners with meaningfully different views on a single important question. No consensus required, debate is the purpose. Submit as a panel organizer with your proposed speakers, or as an individual panelist.

Customer case study + demo

45 mins
Presentation + demo (30) + Q&A (15)

What you built, why you built it, what broke, and what you'd do differently. The most honest format on the agenda and the one that resonates most with a practitioner audience. Production metrics required. Demo required.

Speaker profile

What makes your submission stand out?

We select sessions based on technical merit, practitioner relevance, and originality—not company size, title, or speaking experience. First-time speakers with a strong story are encouraged to apply.

You've built or operated something at production scale

The expertise in your submission should come from direct experience, not research or theory.

Your session is 300-level or above

No introductory or survey content. Attendees are practitioners. They've read the blog post. Give them the insights the blog post can't.

You can show it, not just describe it

Every accepted session (except the panel discussion) requires a live demo. If your content can't be demonstrated in a live environment, it may not be the right fit for this format.

You have a customer story or customer co-speaker

All sessions should include a customer speaker or a referenced customer deployment is possible. Internal-only stories default to a lower priority.

Topics of interest

What we're coming together to talk about

If you're working in any of these areas, we want to hear from you. These are the conversations we're building the agenda around.

Operate AI at scale

Compute and networking at AI scale
Capacity planning and consumption models
Securing AI infrastructure
Operating reliable research clusters across your entire AI infra footprint, including on premises
TCO, utilization, and cost control
Demystifying benchmarks
Real-world AI benchmarking
AI-native storage built for training-scale throughput
Cross-cloud data movement with CAIOS and LOTA
Matching workloads to the right compute (CPUs, GPUs, accelerators).

Make models better

Optimizing training at scale
RL and Serverless RL
Experiment tracking as the bottleneck
Distributed training playbooks
Iteration speed and experiment tracking with W&B
Frontier training lessons learned
Training and post-training economics
Data pipelines and storage for training at scale

Build agents that ship

Monitoring agent fleets and observability
Autoresearch agents
From prompting to systems
Agent protocols and frameworks
Agent architecture, memory, and planning
Multi-agent systems
Evaluation of agents
Sandboxes for agent development
Building eval datasets from production singles
Inference patterns for agentic workloads
Techniques for making the most of your infra footprint

Industry pioneers

Physical AI and robotics
AI in media and entertainment
Quantitative finance and HFT
Life sciences and biotech
Real-time and latency-bound workloads
Line-of-business outcomes and ROI
Customer journeys from pilot to production
Real transformation stories, talk about the things that DIDN’T WORK without feeding into the myths that all POCs fail

Submission Guidelines

What a strong submission looks like

Before you hit submit, here's what separates the sessions that get selected from the ones that don't

Title

80 character limit.

Specific and declarative, quantitative and bold. "How we cut inference cost 60% in 90 days" beats "Optimizing AI inference." Your title is what gets someone to walk into your session instead of the one next door.

Abstract

150 word limit. Answer these four questions:

What problem does this session address?
What will attendees learn that they can apply?
What makes your perspective credible?
What will the demo show?

Avoid vague language about "exploring" or "discussing". Be specific about the value the audience will walk away with.

Difficulty level

All sessions must be 300-level (intermediate-advanced) or 400-level (advanced/expert). We do not accept 100 or 200-level introductory content. If your session assumes the audience needs to be introduced to the concept, it's not flying at the right level for Fully Connected.

Demo requirement

Every accepted session—with the exception of the panel discussion—must include a live demo. Your submission should describe what the demo will show and what environment it will run in. We’re not asking for slides—if your session cannot support a live demo, the panel or workshop format may be more appropriate.

Submit now

Deadlines and Dates

Call for Speakers Deadlines

MAY 2026

Commencement

Call for Speaker Proposals Opens

JuN 2026

Deadline

Deadline for Speaker Proposals

Jul 2026

MIlestone

Notification of Acceptance

Please Note:

We sincerely appreciate the time and thought that goes into each submission and thank you for your interest in contributing to Fully Connected 2026. We review every submission carefully. Due to volume, we're only able to follow up directly with accepted speakers.

Speaker Perks

What's in it for you

Accepted speakers get more than a slot on the agenda. We’ve worked hard to make this worth your time and effort.

Complimentary General conference pass

Full access to all 3 days: keynotes, all 4 breakout tracks, hands-on labs, expo, evening receptions, and the headliner concert.

Professional session recording

Your session is recorded and distributed to the Fully Connected community post-event, extending your reach well beyond the 2,000+ attendees in the room.

Speaker prep support

Dedicated content kickoff call with the Fully Connected team, review cycles for your content and demo, and full A/V technical support on site.

Speaker community access

Join the Fully Connected speaker network and be considered for future CoreWeave and W&B event opportunities.

Promotional amplification

Accepted speakers are featured on the Fully Connected website, in event email campaigns, and across CoreWeave and W&B social channels.

mark your calendars

Save the date

Three days. Four breakout tracks. Two thousand practitioners. Fully Connected is Sep 29–Oct 1, 2026 in San Francisco. We’re built for this.

When:

September 29 - October 1, 2026

WHERE:

Moscone South · San Francisco, CA

Add to calendar

Inclusivity & Community

Built for everyone building AI

Fully Connected is committed to building a speaker program that reflects the full diversity of the AI community—across background, experience level, company size, geography, and identity. We actively encourage submissions from first-time speakers, independent practitioners, and researchers from underrepresented communities. Technical merit and practitioner insight are the only criteria that determine acceptance. All speakers and attendees are expected to uphold our community standards.

Submit Application

Claim your spot on the stage

Fill out the form below. Due to volume, we're only able to follow up directly with accepted speakers.