Artificial Intelligence & Machine Learning

Secure Solutions for the Public Sector

Certified expertise for secure AI/ML deployment

Probity is a certified Google AI/ML Partner with proven expertise in design, development, and deployment of secure systems for the U.S. Government.

ai-ml-expertise
gc-co-sell-partner
gc-services-partner

Why choose Probity for sensitive AI/ML operations?

icon-award

Certified Google AI/ML Partner with access to their latest technology

icon-security

Fully Cleared Experts in Networking, Cloud, and Secure System Design

icon-certified

Original AI/ML research focused on operational problems critical to national security

tech

Deep experience with major Cloud providers

Solution Spotlight -
Deep Bucket Query (DBQ)

A Probity AI/ML solution for finding information in massive stores of private and sensitive documents. Users interact directly with an AI agent.

DBQ offers these groundbreaking capabilities:
  • Search for an abstract concept, not just keywords

    Find items lacking the keywords used to describe the concept

  • Retrieve concise summaries with links to source documents

    Retrieve answers, not documents

  • Chat with your results to drill into them

    Use natural language to refine and narrow your search

  • Set persistent queries to catch new incoming documents

    Get alerts when they match your needs

  • Secure all of this in your own Virtual Private Cloud (VPC)

    Your data never leaves your perimeter

DBQ is based on a technology stack from

logo-google-2x
icon-arrow
icon-google-cloud
Google Cloud
icon-compute-engine
Compute Engine
icon-cloud-storage
Cloud Storage
icon-pub-sub
Pub/Sub
icon-cloud-run
Cloud Run
icon-vertex-ai
Vertex AI
icon-gemini
Gemini
icon-big-query
BigQuery

Operational research built for rapid adaptation

AI/ML original research at Probity is focused on rapid adaptation to streams of information boosted by human operator feedback.  We have extraordinary expertise with audio/video media streams and Human Language Technology..

What does this mean for you?
  • Avoid seeing the same errors repeatedly

  • Begin learning a new task immediately; don’t wait to collect a large training set

  • Avoid the bias of a massive pre-trained model; focus only on your operational domain

Recent publications

A Unified Metric for Simultaneous Evaluation of Error Rate and Annotation Cost

IEEE International Conference on Acoustics and Speech Signal Processing, 2025.

Read Article

Iterative Feedback in the Online Active Learning Paradigm

IEEE Automatic Speech Recognition and Understanding Workshop, 2025.

Read Article

The Value of Corrective Feedback in the Online Active Learning Paradigm

IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026.

Read Article