in your

Observe and refine AI applications by seamlessly adding custom monitors

Backed by

UpTrain Features

Track Model Performance in Real-time

UpTrain observes the performance of your model and pin-points (any) dip in model’s accuracy to a specific feature combination

UpTrain’s Dashboard highlighting the performance of Machine learning model post deployment

Track Data Distribution Shifts as they happen

UpTrain compares your production data-points’ distribution against your training dataset and detects out-of-distribution cases

UpTrain’s Dashboard showing Data Distribution Shifts as they happen

Define Signals to Collect Edge-cases

UpTrain allows you to define “Signals” as a way to incorporate your domain-specific knowledge which are used to filter out specific subsets of data which can be used to retrain the model

UpTrain’s Dashboard showing edge case collection

Smart and Automated Model Retraining

With UpTrain, you can attach your existing data annotation, model training, and deployment pipelines to activate a completely automated continuous model improvement cycle

Pop-up box showcasing retraining upon edge cases collected by UpTrain

Seamless Integration into your existing pipelines

UpTrain offers seamless integration with all the major ML libraries and MLOps tools, allowing you to get started in less than 5 minutes

Image showing UpTrain integration with multiple ML models and cloud providers like AWS, GCP, Azure, Tensorflow, Pytorch etc.

Team

UpTrain is built by passionate ML engineers who aim to leverage their past experience and democratize the best-in-class ML observability and Refinement tooling for the open-source community

Sourabh Agrawal

CEO

Sourabh has built and deployed end to end ML systems of varying complexities at Insane AI (an AI-powered fitness startup he founded), Goldman Sachs (ML in finance) and Bosch (Deep learning consultant for autonomous driving). He is an alum of IIT Bombay.

Shikha Mohanty

CMO

Shikha has been a venture builder and SaaS investor in the past. She has worked with several deeptech companies and has contributed to their growth journeys. She is an alum of The University of Hong Kong.

Vipul Gupta

CTO

Vipul has experience in building large scale ML systems at Meta, Apple and ByteDance. He is an alum of IIT Kanpur and has received his PhD from UC Berkeley, where he worked closely with Professors Kannan Ramchandran (BLISS Lab) and Michael Mahoney (RISE Lab).

What people think about UpTrain

Andre Franca

VP of R&D @ causaLens

I have been looking for a good open-source observability solution for monitoring ML models for a while now and couldn't have been happier to have found UpTrain! Really excited to see what UpTrain team is building and closely following their progress.

David Chang

ML Engineer, Docugami

Model observability is a real pain point for data scientists, and a tool like UpTrain can help identify model failures and their potential causes in order to combat performance degradation ASAP

Ayush Jasuja

Founder, SpeedyBrand

Model observability is one of the key considerations as we look to use foundation models at our organisation and it’s great to see UpTrain building tools which help in monitoring and refining large language models

Sachin Modi

Lead Data scientist, Impact Analytics

Monitoring your models in production is as important as building them in the first place. The AI community lacks good monitoring solutions which allows deep customizations and it's great to see an open-source project like UpTrain solving it head-on.

Harsh Gupta

PhD, University of Illinois at Urbana-Champaign

Adding custom monitors is a critical requirement for any machine learning observability toolkit, and it's great to see that the team behind UpTrain is designing the project with ease of customization at the core of it.

Koulik Khamaru

Assistant Professor of Statistics at Rutgers University

It's great to see an ML observability tool like UpTrain who is combining statistics, deep learning methods, and customization in the right way to create a comprehensive solution that actually works!

magnifying glass focusing on one red pawn

We Are Hiring

We are looking for highly motivated folks to help us build the next generation of ML tooling. We are hiring for ML Engineer, Backend Engineer, Statistics Ninja, Community Manager and Technical Content Writer

Latest from UpTrain AI

Aryan V S | 20th Mar, 2023

An Introductory Guide to Fine-tuning Large Language Models

A "Language Model" is a machine learning model trained to perform well on tasks related to text/language like Classification, Summarization, Translation, Prediction and Generation.

Shikha & Vipul | 14th Mar, 2023

Unlocking the Power of Language Models with UpTrain

If you have connected to the internet in the last 60 days, it wouldn’t be of surprise that you have heard of ChatGPT(or atleast come across the word).

Sourabh & Shri | 7th Mar, 2023

7 Mistakes People Make When Putting Their Models In Production

A critical part of the lifecycle of an ML model is post-production maintenance and performance. Many issues may arise during this period

Tanisha & Vipul | 19th Feb, 2023

Active Learning: Train Models with Less Labeled Data

Most supervised machine learning models require to be trained with large amounts of labeled data. Active Learning can be used to make models smarter with less labeled data.

Vipul Gupta | 21st Jan, 2023

Bias in Recommendation Systems

Modern-day web systems rely on user feedback (such as click activity or ratings) to build ML models to personalize recommendations.

Shikha Mohanty | 14th Jan, 2023

Why do we need to care about Retraining?

The decay in model performance is generally due to reasons like data drifts, label drifts, and concept drifts. To combat these issues, retraining is critical.

Shri & Sourabh | 07th Jan, 2023

5 Great Statistical Methods For Data Drift Detection

As we now know the significance of handling data distribution drifts, the first step of the process is to identify and detect these shifts.

Your Open-source toolkit to refine ML models