At a Glance
| COMPANY | Together with Kai, an AI companion and community app for connection and loneliness, built by a four-person team led by co-founder Neill Kramer and CEO Roshini Tribhuvan |
| SCALE | 302 user-created Spaces and 41,000+ persistent user memory profiles to date |
| USE CASE | Integrated app combining live audio-first matching with mutual-consent video, user communities, an emotion-tagged social feed, and a private AI companion for reflection |
| STREAM PRODUCT | Chat, Video/Audio, Activity Feeds, and AI Moderation (text) |
The Challenge: Real-Time Connection on a Bootstrapped Budget
Together with Kai, an integrated app for companionship, started as a much simpler idea.
Co-founder Neill Kramer, coming out of years in Silicon Valley, began building on the vibe-coding platform Lovable. He entered their Shipped competition, a six-week program offering mentorship and software credits, with an app called Vibe Align.
That early version matched people through a long, detailed questionnaire designed to surface compatibility. But, it didn't go as planned.
"People don't like to answer questions. So, when I met Roshini and we redid it as Kai, we went directly to Stream to get real-time communication going."
- Neill Kramer, Co-Founder
Rebuilding around live human connection meant the team needed reliable, low-latency audio and video plus chat, something well outside what a small, self-funded team could build and operate on their own.
"You can tell Lovable to build chat, but then you have latency issues. And you have to run audio or video yourself. You can't do it without very sophisticated servers and placement all over the planet. We wouldn't have done it."
Chat, Video, and Feeds in Days
With Stream handling the real-time layer, the team's engineer implemented chat, audio, and video in a matter of days.
"It was very quick. Stream was not a blocker."
That speed let the team build out a full set of connected experiences on top of Stream's APIs:
| Kai | A private AI companion for reflection. Its chat interface runs on Stream Chat, while the underlying AI conversation is connected to OpenAI and Supabase. |
| Spark | An audio-first, near-instant match with a random member; if both people opt in, the call becomes video. Built on Stream Video. |
| Spaces | User-created communities; started with two seeded by the team and now grown to 302. Built on Stream Activity Feeds and Video. |
| The Feed | Members tag a post with an emotion before sharing and can browse a Community Pulse of aggregate sentiment. Built on Stream Activity Feeds. |
Then, there's Stream's AI Moderation. For a platform built entirely around people sharing their emotions openly, chat moderation was not optional.
"Your vulnerability is accepted on the platform. We say kindness is mandatory, and that's upheld through Stream."
This combination - a conversational model that lets trust build organically plus moderation that holds a consistent standard - is core to why members stay in the conversation.
Under the Hood
Both halves of Kai's intelligence layer run directly on Stream's Chat and Video infrastructure, combined with OpenAI and Supabase, rather than as a separate data pipeline bolted on afterward.
- Fast brain: scores emotional signal on every message in real time, with no added latency, so a crisis (like a user expressing suicidal ideation) is caught as it happens rather than at a later check-in.
- Slow brain: runs overnight, synthesizing each day's conversations into a durable per-user profile (remembered facts, values, personality) so Kai's tone adapts over time instead of resetting every session.
Spark Drives an 11x Signup Lift
For months after its initial launch, Together with Kai grew slowly as the team experimented with different approaches to bringing people to the platform. The breakthrough came when the team leaned into Spark's core promise of a real person, live, in under a minute.
"We tried many things to engage users. By the time we rolled out Spark - this spontaneous thing of just bumping into somebody on audio or video - we didn't know if it'd get traction. Turns out, our traction went up dramatically."
The results spoke for themselves.
Signup velocity is up 11x since launch, and traffic jumped 7x with the rollout of Spark's matching. Among users who go on to mutually opt into video, day 7 retention is roughly 3x the platform baseline - an early signal that video is a strong deepener of engagement.
What's Next
With the initial surge in hand, the team is exploring Kai Passport, a B2B extension of the same emotional-signal technology behind Kai. It's an aggregate, anonymized mood tracking system for workplace chat communities, positioned as a lighter-weight alternative to traditional employee surveys.
Kramer points to India's CSR mandate, which requires eligible companies to direct 2% of net profit toward public welfare initiatives like health and education, as a potential opening for a workplace wellness offering.
"Imagine a company using chat: we start processing the text, keep usernames anonymous, and give them an aggregate read on department mood in real time instead of a survey."
Turns out, loneliness might not be too hard to solve, with a little help from Stream.