Machine Learning and Data Analytics

Exponent has worked in the artificial intelligence and machine learning (AI/ML) space for 20 years. For government and industry clients, we have investigated how computer agents can improve their perception, cognition, and function from data, knowledge, experience, and interaction. We are uniquely positioned to provide experienced interdisciplinary teams who can expedite key AI/ML objectives:

Modernizing Analytics Programs

Combining expertise in data analytics and machine learning with expertise in diverse engineering and scientific fields, our teams help build and modernize existing analytics programs for government agencies and corporations entering the data science lifecycle at any stage. Our modernization program can include:

  • Building databases with data that were not collected with analytics in mind, including non-digitized data, data with missing fields, human error or bias in collection, inaccuracies, and inconsistent collection procedures.
  • Independently testing and validating AI/ML models using various performance metrics and cross-validation techniques.
  • Augmenting existing data sets with non-traditional data sources, like images, text, and smart wearables, using computer vision, natural language processing, and signal processing techniques.


Exemplar projects: 

  • Digitizing, indexing, merging, cleaning, and making accessible over 70 years of historical data from more than 100 sources, including standard forms, tables of numerical data, handwritten notes, images, and geolocation information for a large utilities client.
  • Augmenting existing classical risk models with predictive maintenance ML models on historical equipment installation and sensor monitoring data sets for a large utilities client.
  • Writing scripts to automate repetitive data processing tasks like renaming files, populating report templates, or extracting text from images.
  • Detecting objects or segmenting images to automate feature extraction instead of cost-prohibitive manual labelling.

Process Automation, Optimization, and Quality Assessment

Our experts provide high-level process overview and recommendations for places to improve efficiency, production, and quality, including:

  • Writing scripts to automate repetitive data processing tasks like renaming files, populating report templates, or extracting text from images.
  • Detecting objects or segmenting images to automate feature extraction instead of cost-prohibitive manual labelling.
  • Automated data quality inspection on a production line using image analysis and computer vision techniques enabling 100% inspection instead of manual subsampling.

Exemplar project:

  • Meta-data analysis to measure diversity of collected data from images of data collection scenes using computer vision techniques. Statistical techniques are used to quantify diversity to guide future data collection efforts for a large consumer electronics client.

Development of AI/ML Tools

With a deep toolkit including machine learning on structured data, computer vision on images and video, natural language processing on textual data, and forecasting time-series and sensor streaming data, our team develops custom AI/ML algorithms.

Exemplar projects:

  • Real-time sensor streaming and fusion of ground penetrating radar (GPR) data to detect IEDs in the battlefield using deep learning models and a custom hardware solution for the U.S. Army.
  • Proposing predictive models for personnel of a government agency using HR analytics techniques to predict risk for attrition.
  • Training computer vision algorithms to recognize key morphological features in SEM and CT images to quantify the amount of corrosion or particle size and count.
  • Running AI on embedded systems for real-time facial recognition and outcome prediction, including health status.
  • Exploratory data analytics to extract trends and make predictions based on sensor data, such as robotic surgical instruments, air quality sensors, and automatic data loggers.
  • Developing predictive maintenance models for field-deployed equipment including historical data and real-time sensor monitoring data for a large utilities client.


Technology Assessment and Regulatory Guidance 

Exponent uses a multi-disciplinary approach to evaluate emerging technologies using AI. Our team of statisticians, machine learning developers, programmers, and cybersecurity experts can assist with advising on regulatory issues (e.g., SOTIF, FDA documentation requirements and software V&V, ISO90003/25000) and help evaluate intellectual property.

Exemplar projects:

  • Evaluation of AI-based in-line manufacturing tools and robotics for regulatory compliance.
  • Software verification and validation for AI-based medical devices and manufacturing tools.
  • Independent cross validation of proprietary AI-based forecasting models for a large marketing company.
  • IP issues related to patent and copyright infringement and product liability for failure of automated systems.

Application of AI in Cybersecurity and Securing AI Data Pipelines

With a strong publication and conference record, numerous professional certifications, and a long history of cybersecurity project support, our teams can help evaluate risk and safeguard data assets in such areas as:

  • Mitigating malicious attacks like poison data injection into an AI/ML pipeline and spoofing autonomous vehicle sensors.
  • Assistance with secure-by-design protocols and implementation, including securing disposable medical device components to protect patient data and client intellectual property.
  • Ensuring security for big data collection effort across U.S. and international cities, project management, logistics, proprietary equipment operation and security, and database management.
  • Risk and threat assessment in the form of preliminary hazards analysis (PHA) and failure mode and effects analysis (FMEA) for security in all systems ranging from supply chain and manufacturing to mobile device security architectures.

Exemplar projects:

  • Performed dozens of quantitative risk assessments using techniques including fault tree analysis, reliability block diagrams, time-based simulations, and Markov models, including several on drive-by-wire vehicle controls systems and hybrid vehicle transmissions.

Artificial Intelligence Risk Management and Safety

Exponent scientists and engineers bring significant experience and expertise to bear on managing risk associated with decisions made by AI systems. Whether working with structured or unstructured learning, Exponent consultants apply expert knowledge to the architecture and design of multiple layers of protection at subsystem and system levels to minimize the impact of undesirable decisions.

Professionals