Antara Dey

Developed an anomaly detection proof-of-concept using unsupervised learning and computer vision to detect foreign objects in factory assembly lines. Configured and fine-tuned anomalib models, resulting in 97% detection accuracy. Streamlined large datasets for class balancing and improving model accuracy.

Setup, tested and documented software of SofDCar seat demonstrator based on Raspberry Pi with CAN communication and compressor pump. Developed automatic speech recognition system in Python using Whisper LLM and Vosk models over REST API commands.

Created high-performance publish/subscribe IPC application using zero-copy shared memory in Rust for video data stream processing. Developed custom software-defined-vehicle Azure IoT and containerised applications in Python/C++ for Yocto-based embedded Linux environment.

Developed and optimized C firmware for low-power basic metering devices in ARM Cortex M through algorithmic improvements, unit tests, bug-fixes, and validation testing. Managed firmware release cycles.

Developed safety-critical firmware for locomotive systems implementing engine management, on-board diagnostics, displays/HMI and compressor control modules in C/C++ on QNX RTOS, cross-compiling for multiple targets and following functional safety standards (EN 50128/50129) for rail industry. Documented complex software architecture using UML.