SigOpt works with algorithmic trading firms who represent over $600B in assets under management. These partnerships have given us unique insights on how model optimization and experimentation can be made most useful for teams who are modeling at scale with the purpose of generating revenue or differentiating their products in a competitive marketplace.
To share these lessons, we are launching a series of talks: Tuning for Systematic Trading. Join this series here.
We know traders know trading better than we do. In fact, given how sensitive these workflows are, we designed our solution so our customers do not need to share any proprietary information. As a result, we do not develop any insights on trading itself. Rather, these engagements allow us to build a body of knowledge related to modeling and tuning best practices. So we will focus this series on a few topic areas that help us share some of these modeling workflow lessons with a broader cross section of the industry.
As mentioned, we will be covering a few different topics during this series. Below is a short summary of each. This content is generally most useful for an experienced – but not necessarily advanced – modeler. If you are just getting started in modeling, we have additional content opportunities with you, so stay tuned to this blog and emails from us to learn more. The focus areas for this series include research insights, modeling best practices and tuning use cases.
Topic 1: Research Insights
The first type of content is research-related insights. SigOpt is unique among software providers in that we have an active research team that explores new optimization and AutoML techniques, consistently writes papers for NeurIPS and ICML, and has developed a patented system for parameter and model tuning. Most of this research focuses on the intersection of model optimization and machine learning model development. More specifically, this research generally focuses on Bayesian optimization and other algorithms that are useful for efficient search for model parameters as well as the best ways to systematically serve these algorithms to scale with enterprise customer needs.
Topic 2: Modeling & Tuning Best Practices
Second is modeling workflow best practices. We work with enterprise customers across six different industries and academic users that span dozens of other domains. This collaboration has given us a unique perspective on best practices for the modeling workflow. Most often this will include a quick demo of specific functionality or a particular technique. But in general we will make this responsive to the most frequently discussed topics with our customers so we can share these lessons more broadly.
Topi 3: Modeling & Tuning Use Cases
The final topic is use cases. Sometimes it is better to show rather than tell. In these presentations we will focus on how we solved a specific modeling problem. These typically include code repos so you can recreate the work. We take customer privacy seriously so these are most often toy rather than specific to an individual user, unless the individual user is presenting their own work.