Ten questions for Sumati, our new AI intern
We have a new team member. Sumati Johri joined Automation Boutique as our AI intern.
Sumati studies Computer Science and Engineering at KIET Deemed to be University in Delhi-NCR, India. At NineAnts, Sumati works with our product team on quality assurance for our AI agent: testing its answers on real treasury questions and helping us sharpen how we evaluate it.
We asked Sumati ten questions.
Welcome to Automation Boutique! Can you introduce yourself?
Thank you for the warm welcome! I'm Sumati, a second-year student in Computer Science and Engineering. I've been programming in Python for three years, and that's how I found my way to AI and machine learning. I taught myself ML algorithms, transformers, RAG, LLM-based applications and agentic AI, and used them in two bigger projects: a Multi-Agent Research Assistant and a Talent Intelligence Engine.
Outside class I take part in hackathons. I won first prize at IIT Delhi's Byte Runner ML Hackathon (Tryst 2026) and reached the semi-final of The Economic Times' national hackathon for AI agents. I also love content writing. I like to bring creativity, curiosity and a sense of ownership to whatever I work on.
What made you choose computer science?
It started at home. My father had a desktop computer, and on Sunday mornings he showed my sister and me how to do basic things on it. I remember playing Solitaire and wondering: how can a whole world fit inside a machine like this? That curiosity never left.
When did you first get interested in AI?
On my first day of college. We got a case study: "AI will replace humans", and had to support or dismiss that claim with examples. Google gave us far more data than we could handle, so I used ChatGPT to structure our results. It came back with a clear document, graphs and annotations included. I wanted to know how it did that. That's where it started.
What's the coolest thing you've built so far?
My Multi-Agent Research Assistant, which I built on my own this summer. It started as a tool that answers questions strictly from one document, for students. From there I worked on vector embeddings, hierarchical chunking, RAG, evaluation tests and reducing hallucinations. It was the first time I combined NLP, deep learning and neural networks in one project, and I put everything I knew into it.
Why an internship at a startup?
At a startup you get to do real engineering: build things, test them, ask questions and see the whole picture. Philip and Jan-Willem described a warm and direct way of working, and that's what I was looking for.
What are you working on at NineAnts?
Quality assurance on the NineAnts agent. I started by reading its documentation: what it should do, which models it uses, and how it's evaluated. Then I tested it with questions, from basic to finance-specific, and shared my findings with the team. I've also suggested ideas for how we test and evaluate the agent, and refined them with the team's feedback.
Treasury and finance are probably new to you. What was your first impression?
They are. Large transactions, multiple currencies and rates, and lots of dependencies between them. It's a big field and I find it fascinating. Working on the agent is a fast way to learn it.
What do you hope to have learned or built by the end of your internship?
How production-grade AI agents are built: what rules and limits they need, how they keep client data safe, and how they should behave. I'd love to build or improve a part of it myself. I believe in owning what I contribute, so that would be very special. 😊
What do you do when you're not studying or coding?
I play traditional card games with my family, read crime thrillers, and I'm always ready with a tip for your next series or movie. I love spicy food. And I've started a journal of funny family stories for my newborn nephew. 😊
What's one thing the team should know about you?
I'm curious, goal-oriented and self-driven. I believe in honesty, ownership and creativity, and I want to bring fresh ideas to the table.
Welcome to the team, Sumati. We're glad you're here.
