OPTaaS: Optimization as a Service

OPTaaS: API for Bayesian Optimization

Please accept all cookies to watch this video.

Request an OPTaaS Trial

powered by Typeform
The Problem: hyper-parameter tuning is expensive

Optimization constitutes an integral part of every business problem and when foregone, can lead to lost revenue, unnecessary costs and an overall drop in competitiveness.

When systems are complex, the optimization of the hyper parameters is either random, manual or foregone as its costs, in both computation and time, are typically high.


OPTaaS aims to make optimization efficient for complex and expensive problems

OPTaaS is a general-purpose Bayesian optimizer which provides optimal hyper-parameter configurations via web-services. It can handle any parameter type and does not need to know the underlying process, models, or data.


The Process

Initialisation: The customer defines the hyper-parameters and their bounds

Optimization cycle:
1. OPTaaS recommends a parameter configuration to the customer
2. The customer evaluates it on their objective function and posts back to OPTaaS the associated score (accuracy, return on investment, Sharpe ratio, etc.)
3. OPTaaS models the correlations between the input parameters and the score to recommend the next configuration
4. The cycle is repeated until the score is optimized


Better Results

Faster Tuning

Cheaper Development

Use Cases

OPTaaS Automatically optimizes:

• Any Black box process
• Data Science pipeline
• Machine Learning/Deep Learning models
• Algorithmic trading strategies
• Financial model calibration

Try OPTaaS:

OPTaaS 2-pager
Intuitions behind Bayesian Optimization with Gaussian Processes

Sign up for a demo now: optaas@mindfoundry.ai


• Unlimited parameters per optimization task
• Integer, numeric, Boolean and categorical parameter types
• Flexible parameter constraints
• Seamless integration via simple API
• OPTaaS never sees your models or data


• Removes the pain of coding with automatic web-services
• Makes your models evergreen cost-effectively
• Reduces the time and cost of identifying optimal parameter configurations
• Frees up valuable time for Data Scientists to focus on extracting insights

Bayesian Optimization for Dynamic Problems
Distributionally Ambiguous Optimization for Batch Bayesian Optimization
Fast Information-theoretic Bayesian optimisation
Raiders of the lost architecture: for bayesian optimization in conditional parameter spaces
Optimization, fast and slow: optimally switching between local and Bayesian optimization

Want to work with us?