AI product development for Comet

comet
Comet empowers data scientists and engineering teams to create more accurate machine learning models that uncover previously hidden insights and drive new business growth.

Stats

3

Vention team members

3

years of active engagement (and counting!)

3

ongoing sub-projects completed

Project description

In the ever-expanding universe of artificial intelligence, Comet shines brightly as an ML platform that helps data scientists and ML practitioners build better AI models — faster. Aware that the pace of AI innovation waits for no one, the company sought a stellar AI development partner to refine its existing products.

Vention proved to be the ideal choice to help Comet introduce new features, integrate with popular ML libraries, and upgrade the user experience.

Experiment management

One of the tasks was to enhance the existing solution in neural network training, equipping users with tools for efficient monitoring and analysis of training processes. We streamlined the collection of essential metrics — now easily displayed on a dashboard that visualizes training progress, resource utilization, and the specific code used. Importantly, this improvement allows engineers to iteratively refine their strategies based on actual results.

What’s more, our team integrated Comet seamlessly with popular Python ML frameworks and libraries to simplify the data extraction process in neural network training. With just one line of code, engineers can now activate automatic data collection and logging, making their projects more efficient.

Plus, we improved Comet's interface to support manual logging of a wide range of data types, such as images and videos. This enhancement has led to more accurate and precise AI models.

Model production monitoring

Next, our role was to elevate Comet MPM, a Python library that enables users to track the inputs and outputs of neural networks in real-time scenarios. Our efforts are concentrated on enhancing the Python component of the solution.

Comet LLM

In response to the surge in large language model (LLM) usage, we helped Comet launch the Comet LLM initiative. It offers a user-friendly API for clearer visualization of interactions with LLMs, simplifying the application debugging process and improving overall AI development.

Impressively, a significant portion of the Python component for this sub-project was developed from scratch by one of Vention’s engineers, which underscores our team's exceptional AI expertise and dedication to pushing the boundaries of innovation and quality.

Results

In the last three years, we've worked hand in hand with Comet, elevating AI development for powerhouses like Uber, Etsy, and Netflix. Notably, even with Comet's extensive in-house expertise, Vention's professionals rose to team lead positions, signaling our high internal standards for professional growth

Equally noteworthy, the Comet LLM project has gained recognition as a trailblazer in the industry. We launched this unparalleled solution among the first — if not the very first — in the world, setting a new standard in LLM development and contributing to Comet's meteoric rise in the market.

comet
Tools & technologies
Backend:

Java

AWS

Frontend:

Javascript

React

SDK

Python

All cases

Contact us