FOUNDER
Who is Agam Jain?

Agam Jain is the co-founder of Tensorfuse, a Y Combinator-backed company that is helping machine learning engineers run serverless GPUs and open-source models on their own cloud accounts.

Agam grew up in India and earned his engineering degree from the Indian Institute of Technology Roorkee, where he was involved in semiconductor device research and developed an interest in machine learning.

After graduating in 2020, he joined Qualcomm and worked on computer vision, publishing papers and patents during his time there.

Agam's interest in entrepreneurship was sparked by his peer network - he saw seniors from his college start successful companies like Razorpay. This fascination, combined with the personal growth he experienced working at the startup Aerotime, motivated him to pursue entrepreneurship.

• “What kept me hooked was the growth that I was seeing in myself, not only just technically, but as a person, like, the ability to handle challenges in my personal life, just by virtue of being at a startup”

• “When I was working at Qualcomm, I had “x” percentage of growth month over month, but when I joined a startup, that growth 10x so that is what has kept me going”

He co-founded Tensorfuse with his long-time friend and co-founder, Samagra Sharma. They had known each other for 10 years, having gone to the same high school and college.

The idea for Tensorfuse came about when Agam and Simagra were building applications using LLMs (Large Language Models) and faced challenges in managing the underlying infrastructure. This led them to develop a solution that abstracts away the infrastructure requirements for machine learning engineers.

COMPANY
What is Tensorfuse?

Tensorfuse is a Y Combinator-backed company that makes it easy for machine learning engineers to experiment with and deploy different models on their own cloud accounts, without having to deal with the complexities of managing the infrastructure.

Agam’s goal with his startup is to become a multi-cloud orchestration layer that simplifies the process of running machine learning workloads across different cloud providers.

Tensorfuse allows developers to:

• Connect: “Connect your cloud account (AWS, GCP, or Azure) and Tensorfuse will automatically provision the resources to manage your infra”

• Deploy: “Deploy ML Models to your own cloud via the the Tensorfuse SDK. Data never leaves your cloud and you can start using an OpenAI compatible API”

• Scale: “Tensorfuse automatically scales in response to the amount of traffic your app receives. Fast cold boots with our optimized container system”

THE PROBLEM

Machine learning and software engineers don't want to manage the underlying infrastructure required to run machine learning models.

They just want to focus on building applications on top of these models, without having to worry about managing GPUs, Kubernetes, auto-scaling, etc.

THE SOLUTION

Tensorfuse provides an infrastructure orchestration layer that eliminates the usual requirements for running machine learning models.

With just a click, users can deploy any arbitrary model from Hugging Face on their own cloud infrastructure in a production-ready setup so it can go directly into production and handle lots of traffic.

This allows the engineers to focus on building applications without having to manage the underlying infrastructure.

OUTREACH STRATEGIES

Agam is focusing on a content-driven strategy to reach developers, as well as an outbound approach to target enterprise customers who fit their ICP.

Organic Content: Tensorfuse publishes a lot of content through their documentation, blogs, and social media posts (e.g., LinkedIn, X) to educate developers about new models, technologies, and how to experiment with them. This content helps drive traffic to their website as developers are interested in learning about the latest advancements in the space.

Outbound Approach: To reach enterprise-level customers, Agam and his team take an outbound approach. They research the target companies, understand what they are doing, and then reach out to the engineering teams, VPs of Engineering, and lead engineers to pitch Tensorfuse's use cases.

IDEAL CUSTOMER PROFILE (ICP)

Companies and teams that are building their own machine learning models, either by fine-tuning or training from scratch.

This is because these customers don't have a large DevOps team to manage the underlying infrastructure, which Tensorfuse provides a solution for.

These customers want to keep their data on their own VPC (Virtual Private Cloud) for security or strategic reasons, rather than using a managed cloud service.

Examples of customers who fit this profile include video, audio, and other companies that are using proprietary software and want a faster, cheaper alternative to solutions like Amazon SageMaker.

WHERE DO YOU SEE TENSORFUSE IN ONE YEAR?

In one year, Agam sees Tensorfuse being a multi-cloud orchestration layer, where users can seamlessly run their machine learning workloads across different cloud providers (AWS, Azure, GCP, etc.) without having to learn the intricacies of each cloud platform.

The goal is to make it easy for users to move their models and code between cloud environments, without the hassle of managing the underlying infrastructure.

Examples:

• “Let's say there is some GPU availability issue right now where you can't get GPUs at your will. If you have workloads that you want to run GPUs on, you can't get them on AWS. We would make it very seamless for you to run it on Azure, or Lambda Labs, or GCP, wherever you want to run it on”

• “Let's say I am running on AWS, and now suddenly I want to move to GCP, I have to learn this whole new world of GCP, right? Or if I want to move to Azure, I have to learn this whole new world of Azure”

• “We provide a uniform layer across all of these. All you need is your code, and we make it portable”

Agam wants to make it very easy and seamless for users to work across clouds.

TL;DR

Agam Jain is the co-founder of Tensorfuse, a startup that provides an infrastructure orchestration layer to help machine learning engineers and developers easily deploy and run their models on their own cloud accounts, without having to manage the underlying infrastructure requirements.

FOLLOW AGAM
I WANT MY OWN NEWSLETTER
I WANT TO BE FEATURED
FOLLOW FOUNDERS BRIEF