Experience
March 2026 – July 2026
Deep Learning Engineer
Temple (Eternal)
Gurugram, Haryana
Built production-oriented deep learning systems for sleep onset/offset detection from wearable physiological time-series (EEG, accelerometer, PPG), with a focus on temporal segmentation, signal decomposition, and real-world hardware constraints.
What I've done
- Developed deep learning models for sleep onset and offset detection using physiological time-series signals (EEG, accelerometer, PPG), applying temporal segmentation and change-point detection to identify sleep/wake boundaries.
- Designed and trained CNN-LSTM hybrid architectures to capture local morphological patterns and long-range temporal dependencies in multi-channel biosignal data for sleep stage characterization.
- Built preprocessing pipelines for raw physiological recordings including artifact rejection, bandpass filtering, and epoch-level feature extraction; applied VMD-based signal decomposition to isolate sleep-relevant frequency components.
- Implemented and benchmarked multiple architectures (threshold baselines, XGBoost, LSTM, Transformer) for sleep onset/offset prediction using epoch-level accuracy, F1-score, and temporal boundary error metrics.
- Collaborated on end-to-end ML pipeline design from raw wearable sensor ingestion to model inference, targeting deployment constraints in consumer sleep-tracking hardware.
PyTorch
CNN-LSTM
Transformers
VMD
EEG / PPG
XGBoost
Time-Series ML