Generative AI is everywhere. Chatbots and virtual assistants respond to customers, Large Language Models (LLMs) help make blog posts, viral videos, and even teach students. But just because it’s widespread doesn’t mean it’s easy to use.
Many businesses struggle to adopt generative AI because of its technical complexity, high costs, a shortage of talent needed to work with it, and scalability problems, among others.
For Kaustav Das Sharma, Founding Engineer at no-code AI platform Gumloop, overcoming these challenges means building AI solutions that are intuitive, scalable, and user-focused so more people can benefit without getting bogged down by the tech.
Kaustav Das Sharma’s Worldly Perspective and Its Influence on AI Development
Sharma’s career path was shaped by a childhood spent moving across the globe. Born in Calcutta, he lived in cities like Milan, Dubai, Stavanger, and London before heading to Montreal for university.
Moving so regularly taught him to adapt quickly and connect with people from all walks of life, but he became aware of the vast discrepancy in quality of life between the places he lived and where he was born.
Witnessing these inequalities at a young age inspired Kaustav to make a difference. Torn between studying economics and engineering, he ultimately chose engineering, not just to build things but to create practical solutions that make life easier for others – especially those without access to resources.
This belief that talent is universal, but opportunity is not, continues to guide his work. From his time at Amazon Web Services (AWS) to his role as the first engineer at Gumloop, Kaustav has consistently designed tools that make AI easier to adopt, giving more people the chance to create and innovate.
At AWS, Kaustav took on a big challenge during the Aurora Hackathon: simplifying how engineers write database queries. Aurora, a tool that helps organizations manage data like tracking sales or adjusting storage, required engineers to use SQL (Structured Query Language), which can be difficult for those that aren’t database experts.
So, Kaustav and his teammate built a generative AI tool that let users describe their data needs in plain language, turning those descriptions into SQL queries automatically. They worked hard to make the tool reliable, cleaning up the data it used and tweaking its accuracy with human feedback and tests.
Ultimately, the tool was easy to use, sped up query writing, and made database tasks more accessible – even for Kaustav, who had limited SQL experience at the time. Their work won the hackathon and became a vital resource for the team.
After this success, Kaustav and his teammate Rahul, co-founder of Gumloop, shifted their focus to education, where they faced a different kind of challenge.
Traditional lectures, where students listen and take notes, haven’t changed much in decades. “Post-COVID and in the mobile era, kids are hyperactive, struggle to focus, and traditional lectures aren’t effective for learning,” Kaustav explains.
To tackle this, they built StudyGuru, a tool that helps students learn in more interactive ways. Users can upload notes, textbooks, or presentations, and StudyGuru turns them into quizzes, summaries, and visual aids. Plus, its AI chatbot answers questions and explains complex concepts.
Kaustav and Rahul then went on to tackle a problem closer to home: how hard it is for people to access mental healthcare, even in countries with lots of resources. They found that therapy often felt out of reach; too expensive, confusing, or frustrating to navigate.
To help, they built Therapai, an AI tool designed to make the process easier by letting users set preferences for the type of therapist they want, whether it’s someone to listen to them or someone more solution-focused. It also has safeguards, like clear consent for data use and safety measures to connect users with real healthcare providers.
Although they couldn’t fully develop Therapai due to time constraints – Rahul was busy with YC and Kaustav was at AWS – the project taught them a lot. It showed the demand for AI tools that are easy to use and designed with the user’s needs in mind.
Kaustav took these insights to his work at Gumloop, where he continues to make technology practical, accessible, and helpful for everyday users.
Kaustav’s 5 Lessons for Practical, Accessible AI
At Gumloop, Kaustav built Gummie, a 24/7 virtual assistant that answers questions and solves workflow problems. Users can describe what they need in plain language, such as “summarize this document and email it to my team,” and Gummie handles the rest. This approach saves time and makes project management easier for people without a technical background.
Gummie has already made Gumloop easier for users, shortening onboarding times and reducing manual customer support. “Imagine speaking your imagination into existence,” Kaustav says. “That’s what Gummie aims to achieve – turning ideas into actions with AI.”
Building generative AI tools like Gummie, StudyGuru, and Therapai has taught Kaustav some important lessons about how to simplify technology for everyday users.
Here are some of his key takeaways:
Lesson 1: Ship Fast, Iterate, and Focus on Users First
When building a new AI product, Kaustav believes the key is to launch a Minimum Viable Product (MVP) – the most basic version of a product that has just enough features to be usable by early customers – quickly. The goal isn’t to have everything perfect right away but to get something into the hands of users, gather feedback, and make improvements based on real-world use.
Lesson 2: Adopt Key Principles from the Start
Kaustav suggests two important principles for any new AI project: separate different services in your code so they’re easier to adapt later, and avoid making decisions that can’t be undone. By moving quickly with decisions that are reversible, you keep your team flexible and can avoid getting stuck in the long run.
Lesson 3: Find Your Community and Showcase Your Work
Kaustav’s career took off when he and Rahul started connecting with the broader AI community. Attending conferences and sharing their work not only gave them valuable feedback but also introduced them to people who would play a key role in their future, like Max, who is now the CEO of Gumloop. “Surrounding yourself with a community of like-minded people can make all the difference,” Kaustav says.
Lesson 4: Think Big to Drive Global Impact
Kaustav believes it’s easy to get stuck focusing on small, niche solutions, but real innovation happens when you think bigger. The goal is to create AI products that can change industries and make a lasting impact on a wide range of users.
Lesson 5: The Role of Feedback in Visionary Design
Kaustav’s earlier projects, like Therapai and StudyGuru, put the user’s needs front and center. As he worked on these tools, listening and implementing users’ feedback allowed him to build with a practical purpose, setting the stage for loftier projects like democratizing AI with Gumloop.
Taking the Gatekeeping Out of Generative AI
While many people are playing around with generative AI in their daily lives, taking it to a larger scale can still feel like something only experts can pull off.
Fortunately, Sharma’s extensive engineering experience and unique upbringing are helping make this technology easier for everyone to use, no matter their background or expertise.
Kaustav Das Sharma envisions Gumloop becoming a platform that transforms how people interact with technology, turning complex workflows into simple, intuitive processes. He aspires to build systems that empower workers in all industries to leverage AI without needing technical expertise.
To learn more about Gummie and Kaustav’s latest projects, check out his Linkedin and visit Gumloop.
