Adityam Ghosh

I’m Adityam Ghosh, CEO and Co-founder of EurekAI.

If I had to point to one moment that started my journey into computer science, it would probably be the first time I watched Iron Man in 10th standard.

I was fascinated by J.A.R.V.I.S. I remember thinking about how cool it would be to have something like that of my own—not just a voice that could answer questions, but something that could actually understand what I was asking and help me with things.

So I tried to build one.

This was around 2014, and my first attempts were, unsurprisingly, pretty basic. I used simple speech modules and a hardcoded list of commands that I used frequently. I could say something, the program would recognize it, and if the command matched something I had programmed, it would do the corresponding thing.

It worked.

But it also became very obvious that this wasn’t J.A.R.V.I.S.

There was no real understanding behind it. It could only do what I had explicitly told it to do. Instead of making me lose interest, that realization made me want to understand why it worked that way and, more importantly, how I could make it better.

That curiosity slowly pulled me deeper into computer science.

Over time, what started as me trying to recreate something from a movie turned into an actual interest in machine learning and artificial intelligence. Eventually, I became a Machine Learning Engineer and got to spend my time working on problems that were much more real—and considerably less cinematic—than building J.A.R.V.I.S.

It was through that work, and later during my Master’s at the University of Auckland, that I ran into a completely different problem.

Research.

I noticed that the difficult part of research wasn’t always finding a paper or some piece of information. There is an enormous amount of research available today. The harder part was figuring out how everything connected.

You might start with one paper, find a reference that seems important, follow that to another paper, discover a different method, look for the implementation, find related work, and eventually end up with dozens of tabs and notes trying to make sense of how you got there.

I experienced this myself, and I also saw the same frustration while working as a Machine Learning Engineer.

The problem, to me, wasn’t necessarily a lack of information. It was the journey between information.

Going from point A to point B often required stitching together a process that was spread across different tools, sources, and workflows. And despite how much technology has changed the way we work, that part of research still felt surprisingly fragmented.

That experience eventually became the reason I decided to start EurekAI.

EurekAI is my attempt to work on a problem that I personally found frustrating. The goal is to make the research journey less fragmented and help people move through research with a little more clarity.

It’s still something I’m learning about myself as I build the company. Being a founder has meant going well beyond the things I was comfortable with as an engineer, and I’m still figuring out a lot of it along the way.

In many ways, though, the motivation feels familiar.

Back when I was in 10th standard, I saw something in a movie and wondered, “Can I build that?”

Today, I’m looking at a different problem and asking a similar question:

“Can I build a better way?”

Outside of Work

I’m not always thinking about technology.

I’m quite into anime, and I spend a fair amount of my free time building Gunpla. I enjoy the process of putting together something piece by piece, especially when a build takes longer than expected and requires a bit of patience.

It’s also a nice change from staring at a screen and thinking about software all day.

So that’s me, more or less: someone who got interested in computers because of a fictional AI, ended up becoming a Machine Learning Engineer, and eventually started a company after running into a problem I couldn’t stop thinking about.

And somewhere along the way, I’m still trying to figure out that J.A.R.V.I.S. thing.

 

Papers & Patents

Below is a record of Adi’s published research and patented systems — spanning collaborative filtering, computer vision on constrained hardware, and AI-driven software performance management.

Papers

• Building a Movie Recommendation System using SVD Algorithm
  International Journal of Computer Sciences and Engineering, Vol. 6, No. 11, pp. 727–729 (2018)
  Co-authored with A. Nath and A. Mitra. This paper presents a movie recommendation system built using Singular Value Decomposition (SVD), a matrix factorization technique used in collaborative filtering.
• Bridging Visual Representation and Efficiency for Resource-Constrained Devices
  2024 IEEE International Conference on Artificial Intelligence in Engineering
  Co-authored with X. Ye and S. Manoharan. This paper addresses how visual representation models can be made efficient enough to run on resource-constrained (low-power, low-compute) devices.

Patents

• Recommending Remediation Actions for Incidents Identified by Performance Management Systems
  US Patent 11,860,726 (2024)
  Co-invented with A. Mandal, P.S.K. Reddy, S. Shetty, S. Merin, R.S. Mulpuri, and H. Zhang. Covers a system for automatically recommending fixes for incidents flagged by software performance management tools.
• Forecasting of Resource Requirements for Components of Software Applications
  US Patent Application 18/062,033 (2023)
  Co-invented with R.S. Mulpuri, A. Mandal, and P.S.K. Reddy. Describes a method for predicting the compute resources software components will require.

Full publication and citation record available on Adityam Ghosh’s Google Scholar