Services

WHAT WE OFFER        USE CASES        METHODOLOGY

What We Offer

Our deep neural nets (patent-pending Lt-Wt Nets) offer unique benefits at the edge as well as in the cloud through our EDGXL and CLDXL solutions.

EDGXL cuts down on the frequency and amount of transmitted data by transforming high-volume raw data into much more compact insights. It does that on low-end hardware while operating on µWs of battery power.

CLDXL translates the insights received from a multitude of edge devices into decisions rapidly while employing very few resources. This results in a low latency, high throughput system.

EDGXL can add intelligence to edge devices with the help of a low-end MCU/FPGA while utilizing only a few kB of memory. The EDGXL inference can be implemented on a Cortex-M4 with only 0.1 kB of code. It can perform, for example, Human Activity Recognition with only 2.5 kB of memory, while consuming less than 1 µW.

Use Cases

  • Malware detection
  • Fault prediction in safety-critical automotive subsystems
  • Replacement of large layers of trained conventional deep nets with accelerated/low-memory versions
  • Remote environmental sensors
  • Fall-prediction wearable for seniors
  • Running efficiency indicator for marathon runners
  • Smart drip irrigation
  • Predictive maintenance
  • Disposable, in-body drug-delivery devices

XADD Methodology

We employ the XADD methodology to create measurable business value, rapidly. In most cases, we aim to generate deployable solutions within 30 days of the project kickoff meeting.

  • Agree on business goals, metrics
  • Identify end-user, sysadmin
  • Identify data needs
  • Explore existing data
  • Identify needed, new data
  • Identify speed, energy, memory constraints
  • Identify deployment environment
  • Identify training needs
  • Analyze existing data for quality
  • Facilitate new data collection
  • Clean, transform data
  • Select machine learning models
  • Communicate expected results
  • Establish prototype evaluation protocol
  • Agree on installation, user manuals
  • Develop machine learning models
  • Communicate expected results
  • Prototype; get feedback; repeat
  • Build system from final prototype
  • Deploy the solution
  • Train sysadmin, end-user
  • Establish support protocol