Striveworks Solutions for the Public Sector

  • Chariot

    Chariot, an end-to-end MLOps platform, supports all phases of mission-relevant analytics: model development, deployment, monitoring, and remediation. The platform supports custom, third-party, GOTS, and COTS models and is tailored to the needs of highly regulated and highly secure environments.

    Ease of use, speed, scalability, and auditability are the pillars of Chariot. Its no-/low-code solution enables users of all skill levels to build better models, faster. Chariot will allow your team to develop, train, deploy, monitor, manage, retrain, and redeploy as many custom models and custom workflows as required.

    Chariot enables your agency to embed structured, scalable model development and management inside your existing workflows. The 'Process as Code' and Data Lineage system enables governance and audit on the entire analytics lifecycle.

    Striveworks’ pioneering work in delivering data science and software solutions to DoD customers inspired the creation of Chariot, a “factory floor” capability where numerous disparate and low-code users can engage and draw upon the scalability of an MLOps platform. Chariot is purpose-built from the needs discovered through experience delivering AI/ML solutions in operational environments.

    Chariot’s features allow organizations to address the unique challenges of operational data science in a cost-effective and scalable manner:

    1. Model-in-the-Loop Annotation: Chariot integrates AI models into the annotation and training framework. This integration accelerates model development by reducing the cost and time associated with the typically lengthy training dataset annotation processes. Users are presented with model-suggested annotations, and the changes and edits made to the suggested annotation are incorporated into subsequent model training. Reducing the annotation process to what is often a simple ‘yes’ or ‘no’ selection accelerates the data labeling significantly.
    2. Model Provenance and Audit: Chariot provides a deep audit capability through its patent-pending Chariot Data Lineage (CDL) subsystem. The CDL logs inputs and outputs from every Chariot microservice and tracks deployed versions of workflows and models. Any recommendation, inference, or workflow output generated on the platform can be recalled for inspection to determine exactly how an outcome was derived.
    3. Training Scheduler: The scheduler in Chariot offers efficiency for model training using a low- to no-code environment. Chariot automatically assigns training jobs to available compute—on-premises and in the cloud—and reports the status of runs. This allows data scientists to fully abstract multi-user/multi-cluster infrastructures.
    4. Microservice Architecture: Building from the flexible Kubernetes infrastructure, Chariot can handle the following microservices on behalf of users: software installation, authentication/authorization, logging, metrics insights, security, service discovery, and CI/CD.
    5. Model Catalog – Chariot’s model catalog provides a repository and catalog of AI/ML models. The model catalog lists available models, organizes models by project, and displays metadata for each model. It records, manages, and displays configuration information for each model, and provides the ability to export a model from the model catalog into a standard format (e.g., ONNX), for use in other tools and platforms.
    6. Serverless Inference Framework – Chariot provides a serverless inference framework for the deployment of AI/ML models from the model catalog. This sets up a REST API endpoint for sending data to a model and getting back its predictions. The service is backed by either CPU or GPU compute and automatically scales depending on the amount of data getting sent to a model.
    7. Workflow Engine – Chariot’s workflow engine enables distributed data pipelines and workflows and supports multi-tenant architecture. The workflow engine is positioned to support streaming data in addition to batching.

    Watch this video to learn more >>

  • Ark

    Ark is an edge model deployment software for the rapid and custom integration of computer vision, sensors, and telemetry data collection. This software was initially built to meet operational requirements for stay-behind devices that could operate without requiring heavy RF/data backhaul pipes. Ark provides a scalable means to quickly integrate new sensors as mission requirements change. Ark also provides the Command-and-Control function to manage a fleet of sensors collecting data (e.g., on multiple UGVs or UMSs) while still allowing custom models—developed in Chariot or elsewhere—to be deployed to an edge sensor either individually or as a managed fleet of sensors.

    Ark will allow an edge device to continually process its environment and push new analytics to the edge, reducing bandwidth and response time. The system data flow has many attachment points, which allows for functionality to be expanded quickly. This framework maximizes agility and supports many use cases. Common use cases are built in, such as the ability to alert a user when a target is spotted. Operational data collected by Ark can also be used to further refine and retrain ML capabilities in Chariot or other systems.

    Watch this video to learn more >>

  • Professional Data Teams

    Striveworks’ Tactical Data Exploitation Team (“TDET” or “Data Team”) have delivered a wide range of software applications in operational settings using a common containerized application stack for easier deployment and management on government cloud and on-prem infrastructure. For one DoD customer, over the course of 24 months, the firm’s 2-person TDET completed twenty-five custom AI/ML capabilities defined by end-user requirements, saving more than 35,000 labor hours—and counting—in process automation solutions.

    A selection of these capabilities include:

    • Developing AI/ML Workflows for Sensor Fusion
    • Structuring Data and Creating Cross-Domain Tools for Indexing and Exploitation of Intelligence Holdings
    • Building Computer Vision and Data Tools for Triage of Captured Enemy Media
    • Structuring Biometric Information Data
    • Making Archived Situational Report Data Discoverable and Searchable
    • Detecting OSINT Sentiment