Speaking & conference strategy

How to Get on Stage at AI Conferences

10 min read · Updated April 2026 · Free playbook
Power tip
Conference organizers don't want another "Introduction to LLMs" talk. They want specific, experience-based stories: "How we reduced inference costs by 60% at [company type]" or "What we learned deploying RAG in regulated healthcare." Specificity is what gets you selected from hundreds of submissions.

Speaking at conferences is one of the highest-leverage career activities for AI professionals. A single well-received talk can generate dozens of LinkedIn connections, multiple job inquiries, and invitations to speak at other events. It's also one of the most intimidating — which is exactly why it's valuable. The bar to entry feels high, but the actual path is surprisingly systematic.

The Speaking Ladder

Nobody goes from zero speaking experience to a NeurIPS keynote. The path is a progression, and each level builds the credibility and skills for the next:

Level 1 — Internal presentations: Present at your team's knowledge-sharing sessions, lunch-and-learns, or all-hands meetings. This is zero-risk practice. You learn to structure a narrative, handle questions, and manage time. Do 3-5 of these before going external.

Level 2 — Local meetups: Present at your city's Python, data science, or ML meetup. Meetup organizers are always looking for speakers — many struggle to fill their speaking slots. A 20-minute talk at a local meetup is the perfect first external speaking engagement.

Level 3 — Regional conferences and company events: Apply to speak at regional tech conferences, company-sponsored events, and virtual conferences. These have moderate competition (50-200 CFP submissions) and give you a real stage experience with professional production.

Level 4 — Major industry conferences: Apply to speak at events like PyData, MLConf, AI Summit, or industry-specific AI conferences. Competition is higher (300-1,000 CFP submissions) but your previous speaking experience gives you a significant edge.

Level 5 — Keynotes and invited talks: These come from reputation, not applications. Once you've spoken at 5-10 conferences and built a body of public content, organizers start reaching out to you. This typically takes 2-3 years of consistent speaking.

Building Your Speaking Portfolio

Before applying to any conference, you need evidence that you can deliver a good talk. Build this portfolio proactively:

Record your talks: Even internal presentations. A 15-minute recording of you presenting clearly on a technical topic is powerful social proof for CFP reviewers. Upload to YouTube (unlisted is fine) and link in your applications.

Create a speaker page: A simple page on your personal site listing your previous talks with titles, venues, and links to recordings or slides. This takes 30 minutes to create and immediately separates you from applicants who have no speaking history.

Publish your slides: Share presentation slides on Speaker Deck or LinkedIn after every talk. This extends the reach of your talk and creates searchable content that conference organizers find when looking for speakers.

Finding Speaking Opportunities

CFP aggregators: Sites like Sessionize, PaperCall, and Confs.tech list open calls for papers. Set up alerts for AI/ML/Data keywords and apply to 5-10 per quarter. The acceptance rate for well-written CFPs is 15-25% — volume matters.

Direct outreach to organizers: Follow conference organizers on LinkedIn and Twitter. Engage with their content. When CFPs open, you're not a stranger. Some organizers also accept direct pitches outside of the formal CFP process — a personalized email with your topic, your credentials, and a link to a previous talk recording can work.

Company-sponsored events: Many tech companies host public events (AWS re:Invent community talks, Google Cloud summits, Microsoft Build sessions). These are excellent venues because they have large audiences, professional production, and the association with a known brand adds to your credibility.

The Talk That Gets Accepted

Conference reviewers see hundreds of generic talk proposals. What stands out: specificity (not "Machine Learning Best Practices" but "How We Reduced False Positives by 40% in Production Fraud Detection"), practical takeaways (attendees should be able to apply at least one thing from your talk on Monday morning), and narrative arc (problem → approach → unexpected challenge → solution → lessons learned).

The best conference talks are case studies from real work. Strip out proprietary details, focus on the technical decisions and their outcomes, and your talk will be both unique and valuable. Nobody else can give your talk — that's your competitive advantage.

← Back to Career Navigation