Signal processing, on-device ML, and evaluation systems.
Software R&D and prototype integration for defense and dual-use applications. Work spans acoustic sensing, computer vision, and the systems that retain and evaluate their results.
Capabilities and deliverables
- Sensing and inference
Acoustic event detection, on-device model integration, image registration, and geometric measurement.
Deliverables: Instrumented prototype, runtime integration, and profiling report.
- Technical evaluation
Reproducible comparisons, reviewed datasets, latency measurement, and failure analysis.
Deliverables: Test plan, evaluation harness, and report with operating limits.
- Application and system integration
Device and web applications, review interfaces, asynchronous synchronization, and data provenance.
Deliverables: Working implementation, interface documentation, and repeatable validation.
Commercial implementation
WhirledPiece
Version 1.0 released September 9, 2026. Available now on the App Store.
Measure impulsive acoustic events on a phone while retaining a correctable record under range noise and reverberation. Derive paper-target practice scores from photographs with reviewable alignment and marks.
A signal-processing frontend locates the acoustic onset. A separate on-device classifier accepts, flags, or rejects candidates while preserving the onset timestamp. The image path maps a target photograph into physical coordinates before scoring. Device and web interfaces retain corrections and model provenance.
| Processing stage | p50 | p95 |
|---|---|---|
| Onset-localization processing | 1.37 ms | 2.14 ms |
| Classifier pass, five windows | 65.9 ms | 111.8 ms |
Published production telemetry snapshot from an inventory of 495 sessions, concentrated in one shooter's devices and ranges. These are stage execution times, not timestamp-error or detection-accuracy estimates. Results do not establish performance in other environments. Method and measurement context.
The implementation supports work on event timing, training instrumentation, visual measurement, and evaluation workflows. Adaptation to a new sensor or operating environment requires separate integration and validation.
Execution and data boundaries
- WhirledPiece
- Acoustic inference runs on the device. Saved recordings, target photos, and review data synchronize to cloud services when connected. See the product privacy policy.
- Soundside.ai
- A separate hosted platform for media generation and processing APIs. Operations may use external model providers. See the platform privacy terms.
- Project work
- Execution environment, data flows, external services, and acceptance criteria are specified for the engagement. Commercial product configurations do not define a customer project's deployment.
Contracting and technical inquiries
Feasibility studies, bounded prototypes, evaluation work, and software integration. Deliverables, execution environment, data handling, and acceptance criteria are defined per engagement.
Soundside Design LLC
Fairfield, Connecticut
Registered in SAM.gov. UEI and CAGE available on request.
NAICS alignment: 541715 (Physical and engineering R&D); 541511 (Custom computer programming); 513210 (Software publishing). Registration details available on request.