Deep learning scientist, last startup acquired by Snap to build My AI chatbot, is seeding its latest venture, a platform to build and operate real-time video-based conversational AI agents We have raised funds.
Today, the startup known as eSelf is coming out of stealth with $4.5 million in its coffers and joining the race to build live, interactive, and more realistic AI assistants. The initial focus will be on areas such as education, sales, financial services, real estate, and health and wellness.
The initial pitch for eSelf was that agents currently have query/interaction response times of less than 1.7 seconds. The company claims this is faster and therefore more realistic than voice responses from leading companies such as OpenAI. An Israeli startup has launched a “no-code” creation studio that allows customers to build their own agents. These agents can interact with popular apps like Calendly and Salesforce to schedule appointments, access accompanying content, and view data just like a human assistant.
eSelf says it has already quietly enabled “millions” of real-time conversations. Current clients include Christie's Real Estate and Brazilian bank AGI.
Portuguese company Explorer Investments led the round, with participation from Ridge Ventures and strategic angel investors including Eyal Manor, former VP of Engineering at YouTube and Chief Product and Engineering Officer at Twilio.
eSelf was co-founded by Dr. Alan Bekker (CEO) and Eylon Shoshan (CTO), and its origin story is an unlikely one.
Becker, who lives in Tel Aviv, said his original desire was to become a rabbi. He said he became obsessed with that ambition a few years into his studies, but only after approaching the school's chief rabbi and asking him big questions about God. And he wasn't satisfied with the answer he got.
“I went home to my wife Lisa,” he said. “I told her I no longer believed in God.”
I don't know if he eventually reconciled it. But in the wake of that dramatic moment, he seems to have discovered a different kind of religion.
“My passion has always been in understanding humans and our world in general,” he said. Having had a long-standing interest in mathematics since his student days, he dropped out of rabbinical school and turned to university to study engineering, physics, and computer science.
He enrolled in a machine learning course “completely by mistake,” but it gave him a glimpse of a different kind of omniscience, he said.
“Mathematics allows us to understand natural language, vision and sound,” he said. “It was shocking.”
First startup sold for $100 million
After completing a PhD, many research papers, and a large loan (currently being repaid), Becker has become a specialist in speech recognition, neural networks, and deep learning. It turned out he also had an entrepreneurial spirit and a knack for seeing the proverbial direction of the pack.
His first startup was Voca, which focused on building AI voice assistants for call centers. We started getting big customers like American Express and TNT (now part of FedEx), and then along came Snap. As we reported at the time, Snap acquired the startup for $100 million in 2020, with little indication of how it would use its technology.
Becker took on the role of Head of Conversational AI at Snap, where he led the team building conversational AI technology. But it took almost three years for it to finally find its way into a product, with the viral arrival of generative AI forever changing the world of technology.
When Snapchat announced My AI, a conversational chatbot, in 2023, Becker had been out of the company for a year. Perhaps similarly, my AI has had mixed (and fairly negative) reactions since release.
(Becker said he was involved in the technology, but not in building the product or integrating it with Snapchat.)
During that time, he started thinking about what the next generation of AI agents would look like. By using text-based responses rather than voice or text-based queries, we saw the potential for more interactive, human-like products using video and audio. response.
Bekker co-founded eSelf in collaboration with Shoshan, a natural language processing expert who cut his teeth in Israel's Unit 8200.
The world of technology is rapidly becoming crowded with companies eyeing what is now referred to as the “agency” opportunity.
A number of companies are emerging that are working to take advantage of advances in generative AI to provide responses that are not only more accurate, but also more reliable, depending on the format in which they are provided (“human” and “voice”). I am.
Companies like H, Eleven Labs, Amazon, Anthropic, the newly launched /dev/agents, LinkedIn, and OpenAI are creating building blocks for creating agents that perform a variety of tasks. Startups and established companies like Salesforce also offer agents for specific use cases such as shopping, supporting enterprise sales and revenue teams, and design.
Although it's still early days, it will be worth seeing which startups get paying users and which get investors to help them grow.
That is, assuming one of these takes off and overcomes some of the key gating factors. One area that has proven to be a major drain on money is model building and training. Bekker said eSelf did not build its own LLM from scratch. Instead, things like the llama model created by Meta have been tweaked. “Customers who have built chatbots based on something like GPT can change to use that, but the latency will be longer,” he said.