CHAPTERS
- 0:05 – 1:01
Recall.ai today: meeting-recording API and business snapshot
David explains Recall.ai as an API that lets developers capture real-time audio/video from Zoom, Teams, Google Meet, phone calls, and in-person conversations. He shares traction metrics—1,000+ customers, nearly $20M in revenue, and a team under 30 people.
- •API for capturing conversation data across major meeting platforms and modalities
- •Infrastructure layer used by 1,000+ companies
- •Nearly $20M in revenue with <30 employees
- •Positioned as developer infrastructure rather than an end-user app
- 1:01 – 2:09
Running conversation capture at massive scale on AWS
The conversation shifts to the technical scale and reliability demands of recording and processing meetings. David describes enormous compute and throughput, emphasizing the engineering intensity of being critical infrastructure.
- •~8 million EC2 instances launched per month
- •Up to ~3 TB/sec of raw video processed at peak load
- •Reliability is existential: missed recordings can’t be recovered
- •Growth: 10× usage over last 18 months with another 10× expected
- 2:09 – 3:51
From 19-year-old hackathon winner to a YC interview
David recounts attending the YC hackathon while a Waterloo student, unexpectedly winning, and getting a guaranteed YC interview. He describes rapidly scrambling to turn a hackathon project into something sellable in just weeks.
- •Flew to Mountain View using saved cash to attend YC hackathon
- •Michael Seibel challenged business viability during presentation
- •Win led to a YC interview on a compressed timeline
- •Rushed to sell $20/month licenses to classmates for early traction
- 3:51 – 4:40
Skipping exams, betting on YC, and getting into W20
David describes the intensity of balancing school with the YC interview, including skipping multiple exams and putting flights on a credit card. The risk paid off when the company was accepted into the Winter 2020 batch.
- •Interview week conflicted with six university exams
- •Flew from Canada to Mountain View for the in-person interview
- •Made a high-risk commitment before having certainty
- •Accepted into YC W20 shortly afterward
- 4:40 – 8:00
Co-founder transitions and the path to Amanda
David discusses working with multiple early co-founders and why those partnerships ended—mainly differences in risk tolerance and commitment amid life changes and COVID. He explains why alignment on purpose and perseverance became the key criterion for choosing a long-term co-founder.
- •First co-founder didn’t want to leave college (values alignment issue)
- •Second co-founder sought stability when COVID disrupted plans
- •David spent months as a solo founder
- •With Amanda: explicit agreement to persist through the ‘natural state’ of startups
- 8:00 – 9:01
Building the first product: a call recorder before LLMs
Recall.ai started as a call recording product designed to make research recordings usable. David explains that the product forced the team to become experts at reliability, scalability, and always-on infrastructure.
- •Built a recorder to handle pages of notes and huge volumes of recordings
- •Pre-LLM era: video conferencing adoption was accelerating
- •70–80% of engineering time went into infrastructure, not by choice
- •Downtime meant permanent loss of critical customer recordings
- 9:01 – 10:01
Pivot: turning hard-won infrastructure into an API for conversation data
As LLM capabilities rapidly improved, David and Amanda recognized that many AI products would need conversation data—and that Recall’s infrastructure could be sold as a platform. They pivoted in early 2022 to offer recording as an API rather than an end-user tool.
- •LLMs unlocked new products on top of unstructured conversation data
- •Recall already had the core infrastructure others would need
- •Pivoted to the API model in February 2022
- •Positioned as foundational plumbing for the emerging AI stack
- 10:01 – 12:24
Fundraising the hard way: deferred Demo Day, then 120 seed meetings
David details a difficult fundraising journey across two rounds: a small raise for the recorder and a larger seed for the pivot. With few connections during COVID, they relied on relentless outbound networking to get introductions and meetings.
- •First raise: ~4 months to raise $160K for the recorder
- •Seed for Recall: ~8–9 months total fundraising
- •120+ investor meetings, largely via founder-to-founder intros
- •~1,000 handwritten emails to founders produced ~10 investments and ~$2.5M
- 12:24 – 13:42
Conviction from living the problem: ‘I was the one on call’
Asked why he believed the infrastructure thesis, David points to lived experience operating the system in production and dealing with customer pain firsthand. He even describes literal nightmares about being paged—reinforcing how real and costly the reliability problem was.
- •Deep conviction came from running production infrastructure for years
- •Founder personally handled on-call and angry customer escalations
- •Scaling forced repeated infrastructure redesigns
- •Nightmares about pager alerts underscored the stress and stakes
- 13:42 – 14:43
First customers by converting competitors—and the flywheel to 1,000+ companies
Recall’s earliest growth came from approaching former competitors and offering to take infrastructure off their plate. Once a few adopted it and moved faster, competitive pressure helped make third-party infrastructure the default choice for the category.
- •Emailed prior competitors as first prospective customers
- •Initial skepticism: some feared competitive tricks
- •Early adopters saw faster product velocity by outsourcing recording
- •Network effects/competitive dynamics helped normalize Recall as standard infra
- 14:43 – 16:47
Sales lessons: outbound-driven early revenue and demystifying sales
David explains how the team learned sales through repetition rather than hiring it away early. He reframes sales as clearly communicating value and helping customers navigate internal procurement, noting the first $2M in revenue came from pure outbound led by Amanda.
- •First ~$2M in revenue came from 100% outbound
- •Amanda led early selling efforts; founders stayed close to customers
- •Sales = explaining value + helping customers through company process
- •Engineers can learn sales; it’s not manipulation or ‘black magic’
- 16:47 – 20:41
Efficiency at scale: $20M ARR with fewer than 30 people
The discussion turns to operating leverage: high revenue per employee and why the company stays lean. David attributes it to an extremely high talent bar, engineers owning product/customer work, and minimizing information loss from handoffs.
- •Lean by design: hire rare ‘full-stack’ high-agency builders
- •Engineers own product decisions, customer conversations, and even deal support
- •Prior founder experience is a major hiring signal (many ex-founders on team)
- •Reducing handoffs preserves nuance and boosts quality/velocity
- 20:41 – 23:27
Why conversation data is the future of AI (and why Recall is the data layer)
David argues conversations contain most of a company’s real context—far more than written docs—and AI needs that context to be effective. He positions Recall as foundational infrastructure for capturing the world’s most important unstructured dataset at work.
- •There are vastly more words spoken at work than exist on the internet
- •~99% of organizational context isn’t written down; it lives in conversations
- •AI, like new employees, needs conversational context to become effective
- •Recall aims to be the capture layer feeding next-gen AI applications
- 23:27 – 24:42
What’s next: expanding capture modes and tooling + hiring high-agency talent
David outlines product expansion beyond meeting bots, including a desktop recording SDK and future phone/mobile capture, plus storage/query/preprocessing. He closes with what the team looks for in hiring: people who thrive under high standards while supporting much larger customers.
- •Desktop recording SDK enables botless capture from a user’s machine
- •Roadmap includes phone call recording, mobile recording, and richer data pipelines
- •Building more tooling: store, query, and preprocess conversation data
- •Hiring across functions, prioritizing high-agency people who want intense ownership
- 24:42 – 26:42
Founder advice: don’t give up—and expect it to take at least a year
David’s core advice is persistence, grounded in his experience learning many skills simultaneously and succeeding only after time. He emphasizes that product success often takes a year or more to evaluate, aligning with YC’s observation that many companies find traction well after Demo Day.
- •Giving up guarantees failure; persistence buys time to learn and iterate
- •Early founders are ‘doing 20 things wrong’—it takes time to fix them
- •Expect at least ~1 year to know if a product can succeed
- •YC perspective: real traction often comes 1–2+ years after the batch
