sign mlops and error analysis platform

about

Universal Compatibility

Comprehensive error analysis across all ML models, including classical, image, video, and LLMs

Advanced Analytical Features

Includes object-by-object comparison, automatic clustering, and prediction difference analysis for deep insights

Seamless Integration & Security

Supports MLFlow API and offers secure cloud or on-premises deployment

User-Friendly Interface

Intuitive design for both technical and non-technical users, facilitating easy interpretation of results

Enhanced Collaboration & Efficiency

Robust access control and privacy features streamline team collaboration and reduce the need for custom tool development

features

MLOps Platform

You have access to a comprehensive set of MLOps tools: experiment tracking, dataset and model registration, metric storage, and visualization. The platform is compatible with MLflow API.

Diff-Tool

The tool allows for comparison of experiments and artifacts associated with those runs. For example, it can compare model performance on a test/debug sample set.

Side-by-Side Comparison & Data Mark-up

It provides the ability to visually compare and assess the performance of different models, supporting markup for quality evaluation of the results.

Error Analysis of Most Popular Models Types

It supports the comparison of various data types: LLMs, images, audio, and video. If model performance evaluations are available, automated analytics for model comparison is provided, including clustering and visualization of relative differences.

team

Stanislav Kirillov

CEO & CTO Inferscope

  • Former Head of Machine Learning Infrastructure at Yandex
  • Key Developer of CatBoost - classical ML library with 10+ mln downloads every month, used in Netflix, Uber, Yandex, CERN and many other companies
  • 10 years in Machine Learning

Egor Podmarev

CPO Inferscope

  • Former Chief Product Officer in Healthcare System
  • Revenue growth from $900K to $2,2M in 5 years, 3rd place safety rating in the country during covid, established new department from scratch and achieved revenue of $420K in 2 years
  • 9 years in product management, product design and operations

price

Starter

Free

  • 1 project
  • Up to 3 collaborators
  • Up to 1 GB of storage
  • Up to 100 experiments
  • Up to 2000 runs
  • Up to 5 hours of Diff-Tool

Professional

Best Deal

To Be Determined

  • Unlimited projects
  • Up to 10 collaborators
  • Up to 100 GB of storage
  • Unlimited experiments
  • Unlimited runs (uses storage quota)
  • Up to 100 hours of Diff-Tool

Enterprise

Upon Request

Everything from "Professional" +

  • On premise Diff-Tool Deployment
  • Custom number of seats
  • Single sign on
  • Dedicated data storage
  • Custom storage plans
  • Support for in-house data storage with DIff-Tool deployed on premise (only metadata stored in cloud)
  • Dedicated support channel