Purpose-Built Edge ML Models For Your Application
Our expert ML team can readily transform time-series datasets to provide turnkey production ML algorithms ready for customers to integrate into their embedded firmware using our straightforward API. Utilizing the power and productivity of SensiML's development tools, custom engineering services can provide working solutions in a fraction of the time and cost that might otherwise be required.
Delivered models can be provided in either linkable libraries for the chosen target platform or as fully self-contained C source code.
Professional Services Engagement Model
The typical custom engineering project takes between several weeks to several months to complete based on largely on the level of data collection complexity. Since each project and its associated requirements are unique, we provide clients with a customized proposal with a schedule, cost, and action plan tailored to their specific requirements. Below is our time-tested engagement process which is adapted to suite specific needs:
Project Planning
- Define requirements
- Establish data collection methodology
- Identify data sources and/or hardware configuration(s) for capture
- Review and update resource allocation and scheduling as necessary
- Tool training as needed to support data collection
Data Collection
- Collect new or import existing time-series sensor data for client application
- Label data for ground truth to train/test ML model
- Stage new data capture in iterative rounds to gain insight into potential methodology flaws or omissions
Model Generation
- Perform initial AutoML runs with SensiML tools and pre-trained templates
- Modify resulting pipeline based on performance analysis and expertise
- Iterate model train/test to optimize for available dataset(s)
- Profile against required latency, power & memory constraints
Firmware Integration
- Deliver Knowledge Pack (ML model code) and API for client firmware integration
- Assist client in firmware integration tasks and testing as required
Model Support
- Scale out to support production model variances (voice localization, SKU differences, etc.)
- Roadmap planning for new capabilities and model enhancements
- Training to support in-house model development by client if/as desired
Schedule to Discuss Your Edge ML Needs With SensiML
Contact us to:
- Learn more about SensiML's custom engineering engagement process
- Obtain referral info on SensiML's network of design and distribution partners
- Discuss current or planned edge AI sensing projects and how SensiML can assist
- Receive a proposal on your next smart edge ML project
