Worked on the chroma-core/chroma repository to implement Apple Silicon MPS acceleration for the OpenCLIP embedding function, targeting improved performance on macOS devices. Addressed a runtime error by ensuring input tensors are consistently moved to the model’s device, which enabled stable execution on Apple hardware. Focused on embedding generation and machine learning, leveraging PyTorch and Python to deliver faster embeddings without altering user-facing APIs. The work expanded deployment options for macOS by maintaining compatibility and validating the solution on Apple Silicon. This approach reduced latency and improved throughput for embedding workloads, delivering measurable business value through enhanced performance and platform support.