Inside Qonfido: Building AI-Powered Mutual Fund Discovery

I've worked in the tech industry for over five years. Somewhere along the way, I realised that I needed to actually save and invest for my future. So I started, but in all honesty, it felt like a game where nobody had defined the rules.

Everyone said the same thing: "Invest in mutual funds. Start early. Time in the market matters." I listened to seniors, parents, and friends. Quite frankly, I even followed the financial gurus and everything they talked about. But nobody explained why mutual funds, or how I could actually pick one. I only looked at returns, and I learned the hard way that high returns today don't mean anything without understanding risks. You can't just pick a fund because it's trending or has a catchy name.

Discovering good funds was frustrating. I kept wishing someone would just tell me, "This one makes sense for you and your goals." But the people around me had similar backgrounds, similar blind spots - and everyone was suggesting the same names.

I tried ChatGPT and other AI tools. The answers sounded smart but were vague, lots of "it depends" and "consider your risk appetite" without actually helping me do anything. I needed something that understood my goals and gave me clear, actionable guidance.

Qonfido?

When I came across the team at Qonfido, something clicked.

The vision felt different: this wasn't just another fintech app, it was about making expert-grade financial insights accessible for regular people like me. Something that understood me, personally, and help me discover products that would be suitable to my needs.

I wanted to be a part of building that.

Can fund discovery be made smarter?

Today, finding suitable funds feels like navigating a maze of industry jargon and sales pitches. With 3000+ funds to choose from and tens of new ones getting launched each month, its a frustrating exercise.

So we are building with an approach that bridges deep human expertise and the smart scalability of modern AI.

Our Qonfido rating model is created by Analytics experts to curate the best 300-odd funds in the industry.

Then we run the best funds through our AI models to make the ratings simple, personal and easy to understand - without the quant jargon. For example, check out our insights on Parag Parikh’s Flexi Cap Fund.

Finally, we offer a conversational interface to make fund discovery even more engaging and smarter - with our AI Powered Fund Discovery

What’s under the hood?

As the Senior AI Engineer, I’m working to build some of the core intelligence behind Qonfido’s product. Here's are a few interesting things that go under the hood:

Multi-Agent Orchestration. When you ask the AI Qo-Pilot a question, it doesn't just query one database or LLM and spit out an answer. We’ve built a system where multiple agents work in parallel, pulling from different data sources, analysing your intent, processing through the appropriate reasoning models, and assembling a coherent response. Think of it like a team of specialists collaborating behind the scenes to guide you smarter.

Intent Detection that works. Users can ask questions in a hundred different ways. "Which fund is good?", "Compare HDFC Mid cap with Axis Mid cap", "Where should I invest 50k for 3 years?"-each of which require completely different handling. We’ve designed a multi-stage intent detection pipeline that figures out what you're really asking and routes it to the right skills pipeline.

Richer Responses, not plain text. Most AI tools return boring walls of text. We wanted something different. We worked on building a large number of purpose-built response designs: fund comparisons, guided suggestions, performance analysis, calculators etc. - which ensure that every response is structured, visually appealing, and easy to digest.

Ah, if only it were so simple

Making AI insights actually useful and insightful.
It's one thing to pull data. It's another to turn it into something a normal person can act on. We've spent countless hours iterating on prompts, skills, and the tools that power our agents. Getting the balance right between comprehensive and overwhelming is harder than it sounds.

Designing for engagement.
Financial content is usually hard to consume. Making it feel approachable, not intimidating, has required us to rethink everything we present - every template, every message type, every skill, went through multiple iterations till it passed our evaluation standards.

Building at the edge.
Honestly, some of this felt way above my pay grade when I started. Multi-agent systems? AI Memory architectures that remember personal context? But we are build it out piece by piece. That's the part I actually enjoy: the problem-solving, the design and build, the "oh wait, I think I can make this work" moments.

What's next for us at Qonfido?

We're still early. We're learning. And we're building Qonfido to be the AI companion that makes managing money simple, personal, and trustworthy, for everyone, at every stage of their journey.

Here’s your front row seat preview to AI powered personal finance.