Unconventional data science, machine learning, and artificial intelligence for unconventional businesses in unconventional fields.
Delightful Research LLC
HI THERE. THIS IS DELIGHTFUL RESEARCH.
Heard about ChatGPT but don’t know how to put it to use for you?
Have some unusual tasks that can’t be automated away with out of the box services on AWS?
Did you hire a data scientist only to find that you need a little more pixie dust to see the AI transformation you want?
At Delighhtful Research, we’re hyper-focused on helping our clients achieve their AI goals.
Services and example projects:
Financial Compliance
Financial regulators are perpetually finding new issues to be concerned about - as soon as you understand one issue, there’s yet another to be controlled and monitored. Suppose you wanted to introduce machine learning - but you don’t have any data with which to train a model? Delightful Research is proud to offer proprietary methods for the training and evaluation of models for which training data does not yet exist, setting your firm up to hit the ground running with automation from day one.
Cybersecurity
Machine learning for cybersecurity typically relies on sequence based models - LSTMs, RNNs, etc. These are great for pattern recognition - but they don’t actually understand what’s going on.
Delightful Research specializes in taking techniques from symbolic natural language processing and extending them to malware analysis.
Objectionable Content
The emergence of Generative AI has made it easier than ever for bad actors to generate objectionable content.
Delightful Research is proud to offer a comprehensive suite of models ready to interdict the foulest of internet scum and villainy.
Chatbots That Actually Work
Have you ever played with ChatGPT, only to figure out that it’s precisely as clever as a precocious thirteen-year-old child?
Delightful Research has been using Attention networks (the precursor to LLMs) since 2018 and understands how to programmatically prompt Large Language Models into the desired behavior.
Private Credit
Classic credit risk models assume a large history of lending data for a given population - but what if you’re working with new products serving new populations? Delightful Research can help.
Furthermore - because business lending is relatively unregulated - Delightful Research is able to use unconventional feature engineering to dramatically outperform standard approaches.
Marketing Analytics
For some reason or another, certain sectors of the economy attract clients who make a hobby out of first party fraud. Competitive behavior between large websites has made tracking harder than ever. Classical methods of marketing analytics were designed under the assumption of honest customers and clearly trackable website traffic. Delightful Research offers bespoke Customer Lifetime Valuation and precise Marketing Mix Modeling for these scenarios.
Activism Support
Delightful Research is a strong supporter of protected First Amendment activities. Sometimes, exercising First Amendment rights can benefit from real-time geospatial support. Delightful Research has produced real time analytical support to ensure the safety of activists engaged in protected First Amendment activity when faced with an aggressive response.
Model Validation And Controls
While it’s one thing for Apple, Amazon, Google, or Microsoft to play with Large Language Models - it’s quite another for those participating in a regulated industry to do so. Delightful Research pioneered the deployment of natural language processing models in scenarios supervised by the Office of the Comptroller of the Currency and Financial Industry Regulatory Authority.