Team
Team
About us
Revalute Machine Learning based systems from a buzz word to a key business feature
Our Features
ReML
- solves business problems
- hardware as a key to success
- result-oriented proven methodology
- clear and transparent communication
with stakeholders
Others
- build ML-models apart
- miss hardware part
- process-oriented methodology
- black-box development process
● build digital transformation plan
● validate & improve existing strategy
● conduct Design Thinking sessions to mine innovative ideas
● focus on potential business effect
● score & range ML initiatives
● estimate their feasibility
● help develop solution plan
Management
● support on hiring and retention processes
● optimize operations for existing teams
● build efficient DS workflow
Methodology
Our Cases
Goals
● decrease product cost without quality loss
● monitor staff performance
Solution
● software: predictive analytics for melting process, optimization engine for accurate alloy amount calculation; web application with dashboards and analytics
Results
● system reduces alloy consumption by 2-5 % depending on technical process
● allows management to control and stimulate technologists
Goals
● develop visual navigation system for parkings
● detect rule violations
Solution
● hardware: LEDs for navigation, custom device with 4 cameras, US sensor, embedded computer
● software: computer vision algorithms for vehicle
detection, classification & parking correctness, plate number recognition, visual navigation system
Results
● system is able to track & navigate multiple vehicles
● accurately detects improper parking, overspeed
● integrated with parking management system
Goals
● build accurate garbage recognition system and implement it to three sorting devices of different scales
● improve recycling process & allow users to make money
Solution
● hardware: 3 high-resolution cameras + lightning, sorting machines
● software: deep neural network and proprietary algorithms for image processing
Results
● precision varies from 94 to 97 % depending on the trash category
● inference time around 100 ms with several cameras