top of page
natural-language-processing.jpeg

Natural Language Processing (NLP)

Suicide ideation prediction for Veteran Care: SoKat Suicide Ideation Engine (VA SSIE)

​

AI Capability: Natural Language Processing | Intent Detection | Responsible AI | Cloud API Integration

​

Problem

The Veterans Crisis Line (VCL) faced 90% false positives from a legacy keyword-based system that identified suicide ideation in Veteran survey responses. The burden overwhelmed staff, reducing time available for real outreach and care.

 

Solution

SoKat developed and fine-tuned the Suicide Ideation Engine (SSIE), an AI-driven NLP model that analyzes the context and intent of Veteran language to accurately flag at-risk communications. The model was integrated via secure API into the VA Enterprise Cloud, built under Human-Centered Design (HCD) principles to ensure clinician trust and usability.

 

Outcomes

  • Reduced false positives from 90% to under 20%

  • Enabled faster, more accurate intervention for at-risk Veterans

  • Lowered burnout for crisis line staff

  • Advanced responsible AI adoption within the VA mission

GSA-AI-1.png
Project: GSA AI – Machine Learning Competition – EULA Challenge

The GSA sought a solution to address the the problem of the lengthy amount of time being spent on reviewing EULAs, which was an average of  7-14 days,  since the EULA terms had to be manually reviewed and processed. SoKat, in response, built an AI solution, which ultimately was awarded 3rd place in the competition. Our AI solution was a web-based, scalable tool for federal government employees that automated their review of EULAs and dramatically reduced EULA processing time. A few notable features of our solution comprise of:

  • Leveraging AI algorithms to label unacceptable terms

  • Text classification model developed using Machine Learning, APIs and front-web application

  • Complete development and testing undertaken in AWS

bottom of page