Research Lab

Machine learning that survives the real world.

Research Lab is Winter Labs' long-horizon bet on machine learning that holds up outside the lab. We study cross-dataset generalization, efficient inference, and edge deployment — building models that stay accurate when the data distribution shifts and small enough to run on a Raspberry Pi. Findings that graduate become the engine inside our products.

Cross-dataset generalizationEfficient / edge inferenceHealthcare & signal MLReproducible experiments

projects in this lab

<100ms

inference targets on edge devices

tracked milestones

lab designation

1 project underway

// Research Lab

HridAI

Research

Cross-dataset ECG arrhythmia classification for the edge.

PythonView source

HridAI studies whether an ECG classifier can stay accurate when the data distribution shifts — training on the MIT-BIH Arrhythmia Database and transferring to PTB-XL — while remaining small enough to deploy on real edge hardware. The target is a single-lead model that runs on a Raspberry Pi 4 / ARM Cortex-A72 with sub-100ms inference and under 150MB of memory, producing a binary risk triage: Low, Monitor, or Refer.

<100ms

inference

<150MB

memory

INT8

quantization