Enabling AI innovation at the Big Tech scale

AgileEngine’s AI engineering, UX/UI design, and software development expertise powers two AI initiatives at an incubator-style division of a FAANG company. In one of these initiatives, we prototyped and supported the implementation of an AI tool that analyzes legal documents in high-stakes IP cases. The second initiative driven by our team repurposes subterranean fiber-optic infrastructure to identify and classify urban activities, including vehicle and pedestrian traffic, construction work, and more.

Hero image

Industries

Artificial intelligence, Machine learning, Big data, Analytics, Enterprise software, Research and development

Services

AI engineering, Data engineering, UX/UI design, Architecture, Prototyping, Feature development

Solutions

AI, Machine learning models, Data pipeline, Predictive analytics, Data visualization, Enterprise application, MVP, Dashboard

Technologies

Figma, Miro, UserTesting.com, Large Language Models (LLMs), Convolutional Neural Networks (CNN), Distributed Acoustic Sensing (DAS)

AI tools that powered our workflow

Copilot, Cursor, ChatGPT

Outcomes
and highlights

  • Over 90% accuracy ensured for the AI classification of urban activities based on seismic data
  • 30x–1000x higher channel counts compared to research models handled by the delivered AI solution to guarantee real-world performance

“I got to present the clickable Figma yesterday in a few internal forums to explain what we are working on as a system, and it’s super neat and easy to use. Thanks for creating the version, super useful!”

– General Manager and Director

Solutions overview

Project overview

AI prototypes and products with real-world impact

The litigation assistant prototype delivered by our Design Studio experts analyzes legal documents and evidence with unprecedented speed, cost efficiency, and precision. Targeting high-stakes IP and patent infringement cases, this solution replaces fragmented, multi-tool legal review processes with a unified LLM-based system optimized for real-world litigator workflows.

Our work with the company also includes a project that adapts Distributed Acoustic Sensing (DAS) techniques used in seafloor exploration to urban environments. The AI systems we implemented for this project classify urban activities with 90%+ accuracy by using fiber optic cables as seismic sensor arrays. Our experts also identified a critical gap between theoretical and real-world data volumes required for analysis, and closed it by enabling 30x–1000x higher resolution in seismic data processing.

Key deliverables

  • Prototype of an AI/LLM tool that shortens the time needed to analyze IP/PO infringement evidence from hours (or, potentially, days) to minutes
  • AI/data pipeline and systems for a solution that classifies urban activities using fiber optic cables as seismic sensor arrays
  • Data preprocessing pipeline scaling to gigabytes per minute, enabling DC offset removal, anti-aliasing, intelligent downsampling, and spatial aggregation
  • Multi-scale Convolutional Neural Network (CNN) architectures that identify events such as footsteps, vehicle movement, jackhammer operation, etc., with 90%+ accuracy
  • Spatial and temporal pooling layers enabling the AI solution to handle 30x–1000x higher channel counts compared to research models
  • Transfer learning strategies enabling pre-training on controlled research data with fine-tuning on downsampled real-world datasets
  • Visualization tools with custom plotting utilities essential for manually verifying event types and discovering characteristic patterns

Technologies

Figma, Miro, UserTesting.com, Large Language Models (LLMs), Convolutional Neural Networks (CNN), Distributed Acoustic Sensing (DAS)

See more success stories

Scroll to Top