For artificial intelligence to be truly intelligent, it needs a reliable memory. An AI that cannot remember last week's discussion, the language you used, or a critical decision from months ago will always remain a superficial tool. This core challenge is now being addressed by innovators globally, including two Indian-origin entrepreneurs based in the United States. They are pioneering a new category of AI products focused not on complex reasoning, but on effectively capturing, organizing, and recalling the countless conversations that define our professional lives.
From Black Holes to Business Meetings: The Genesis of TwinMind
One of these innovators is Daniel George, an alumnus of IIT Bombay who hails from Kochi. He moved to the US in 2015 to pursue a PhD in astrophysics, which he completed in a record-breaking one year. His pioneering work involved applying AI to gravitational-wave astrophysics to detect black holes, research that contributed to a Nobel Prize for his extended team. This credibility led him to Google X, the company's moonshot lab, where he worked on futuristic projects like AI-powered hearing aids and motorized exoskeletons.
However, the inspiration for his startup, TwinMind, co-founded with Sunny Tang and Mahi Karim, came from a more mundane source: the frustration of forgetting details in an always-on work life. While serving as Vice President of Applied AI at JP Morgan, Daniel experimented with an internal tool that listened to meetings, transcribed them, summarized decisions, and even suggested follow-up actions. When shared with friends, one posted about it anonymously on forums like Reddit, sparking immediate interest. "Hundreds of people said they would pay for it," Daniel recalls.
TwinMind is a software-first solution—a smartphone app and browser extension that runs continuously, capturing conversations and building a persistent, searchable memory of everything a user sees, hears, and says. Daniel has used it for eight months, recording every discussion with investors, co-founders, and family. The magic lies not in transcription alone but in how this memory is structured and reused. The system drafts emails, prepares for calls, builds presentations, and suggests tasks by drawing on past context, acting like a personal Jarvis from Iron Man.
Solving the Technical Puzzle: From 140 Languages to On-Device AI
The technical hurdles were significant. TwinMind supports 140 languages, including Indian languages like Malayalam, Tamil, and Marathi, which are often poorly served. Their breakthrough involved using Large Language Models (LLMs) as judges. They scraped audio from YouTube and podcasts, ran clips through multiple AI models, and used an LLM to determine the most accurate transcription, creating a high-quality dataset to train smaller, efficient models.
"By comparing ten different models, you can correct mistakes and create a ground-truth dataset," Daniel explains. These optimized models run on-device to save battery, switching to the cloud only when needed, keeping costs under a dollar per user monthly. Daniel envisions a future where everyone has a private AI memory layer that other applications can consult, tailored to their individual values and life context.
The Hardware Solution: Buddi AI's Wearable Memory Device
If TwinMind represents a software approach, Anith Patel's Buddi AI offers a hardware answer to the same memory problem. Based in San Francisco, Anith, a BTech graduate from SRM University in Chennai, built Buddi around a wearable clip or pendant. His reasoning is straightforward: smartphone apps face operating system restrictions on microphone access, but a dedicated hardware device does not.
Anith's background in IoT, wearables for the visually impaired, and fintech informed Buddi's development. Early prototypes were 3D-printed in his Ahmedabad lab and tested in the US and UK. The device is designed for B2B use, particularly for field sales teams. It continuously captures conversations, transcribes them, generates summaries and action items, and syncs data to an app and enterprise dashboards.
"Sales managers in the US get a real-time overview of what's happening on the ground, without their teams spending hours on documentation," Anith states. The system automatically logs customer interactions into CRMs and calendars, giving managers analytics on team performance and helping salespeople track promises and next steps. While it supports Indian languages, the primary focus is on achieving a very low word error rate for English-speaking customers.
The Buddi device includes onboard storage and Bluetooth, functioning even without a phone connection. Anith's long-term vision is to build an organizational brain for companies, allowing them to predict problems before they escalate. "We want to build a brain for companies," he says, highlighting strategic organizational memory as the ultimate goal.
Together, these two ventures founded by Indian entrepreneurs demonstrate a pivotal shift in AI development. Moving beyond mere conversation, they are building foundational memory layers that promise to make our interactions with technology deeper, more contextual, and genuinely intelligent.
