How it Works
Integrate Lens AI Client in to your edge device using few lines of code
C++ Profiler API for the edge device
Lens AI Modules
Lens AI Monitor
Lens AI Python Profiler
Lens Cpp Profiler
Lens AI Python Profiler - Get the base profiles using the Lens AI Python Profiler during the training
Lens AI Cpp Profiler - Integrate Lens AI Cpp Profiler into your Edge device Inference code
Lens AI Monitor - Monitor the Drift and sample the data where model is most uncertain
Why Lens AI
Put Lens AI to work and make model debugging and retraining faster,
Invest your time where it's needed.
Save money in labelling and
data transfer
Key Monitoring Metrics Computed
on device with low footprint
Model Monitoring in accordance
with EU AI Act
EU Data Privacy compliant Data
Monitoring
Save time on root cause analysis
and debugging
Memory Effciency
Fixed Memory Usage
The data structures used by Lens AI operate with a fixed amount of memory, making them ideal for applications with memory constraints or long-running processes where memory usage must remain predictable and bounded.
Scalability
Unlike classical histograms that might require more memory as more data is processed or as data complexity increases (e.g., higher resolution, multiple channels), these sketches maintain a consistent memory footprint, ensuring efficient scalability.
The space complexity is 𝑂(1/𝜖log(𝜖𝑁)) where as classical logging it is O(N).
Computational Efficiency
Lens AI leverages KLL datastructures that are designed for incremental updates, making them highly efficient for streaming data scenarios.
Amortized logarithmic insertion complexity ensures efficient handling of high-throughput data streams
Time complexity while insertion and query time better than classical logging
