The startup transforms meeting notes with time-saving features


Gil Makleff and Artem Koren develop AI for meeting transcripts, creating time-savers like shareable highlights of text that’s often TL; DR (too long; did not read).

Sembly’s founders conceived the idea after years of working in corporate operational consulting at UMT Consulting Group, which was acquired by Ernst & Young.

“We had a hunch that if AI was applied to these operational conversations and able to make sense of them, the value gains for businesses could be huge,” said Koren, chief product officer at Sembly.

Sembly goes far beyond basic transcription, allowing people to skip meetings and receive speaker highlights and key action items for follow-ups.

The New York startup uses proprietary AI models to transcribe and analyze meetings, turning them into actionable insights. It aims to energize teams who want to focus on getting results rather than spending time compiling notes.

Sembly’s GPU-powered automatic speech recognition AI can be used with popular video calling services such as Zoom, Webex, Microsoft Teams and Google Meet. With just a few clicks on the Sembly site, it can be synced with Outlook or Google calendars or used for ongoing calls via email, web app or the Sembly mobile app.

The service offers market-leading transcription accuracy and AI-powered analytics, including highlights to identify important talking points. It also allows users to focus on meeting speakers and easily share clips of individual passages with team members, improving collaboration.

Sembly, founded in 2019, is a member of the NVIDIA Inception starter program.

Improving speaker tracking with NeMo

One of the weak points that Sembly addresses in transcriptions is something called diarization, or identifying the correct speaker in the text, which can be problematic. The company had tried popular diarization systems from major software makers with negligible results.

Diarization is a key step in the meeting processing pipeline, as many of Sembly’s natural language processing features rely on this text to be properly identified. Its Glance View feature, for example, can identify key meeting topics and who raised them.

Assigning meeting topics to the wrong person throws a wrench in tracking action items.

Leveraging NVIDIA NeMo, an open-source framework for building, training, and fine-tuning GPU-accelerated natural language and speech understanding models, has enabled a significant advance in accuracy.

Using the NeMo conversational AI toolkit for diarization model training, running on NVIDIA A100 GPUs, significantly improved its speaker tracking. Before applying Nemo, he had an 11% diarization error rate. After implementation, its error rate dropped to 5%.

Boosting activity despite meeting fatigue

With a shift to fewer face-to-face meetings and more virtual meetings, companies are looking for ways to counter online meeting fatigue for employees, Koren said. This is important for providing more engaging work experiences, he added.

“There is a concept of ‘meeting tourists’ in large organisations. And that’s one of those things that we hope Sembly will help solve,” he said.

Adopting Semby to easily highlight key points and speakers in shareable transcripts gives workers more time in the day, he said. And leaner operational technologies that help companies stay more focused on key business goals deliver competitive advantages, Koren said.

For those with busy schedules who need to try dancing in between meetings, Sembly can help too. Sembly can be invited to attend a meeting on behalf of the user and come back with a summary and a list of key items, saving time while keeping teams more informed.

“Sometimes I would like to attend two overlapping meetings – with Sembly, now I can,” Koren said.


Comments are closed.