hey there!

How to use Colab TPUs with large datasets (for almost free)My guide on using TPUs in Colab with large datasets. Avaliable as a medium article.

SERS Nanowire Chip and Machine Learning-Enabled Classification of Wild-Type and Antibiotic-Resistant Bacteria at Species and Strain LevelsI worked together with Dr. Dushan Wadduwage from Harvard university to create a Siamese neural network to classify the Raman Spectroscopy obtained by Sathi's SERS Nanowire chip.

Multi-label Classification of Reduced-lead ECGs using an Interpretable Deep Convolutional Neural NetworkThe model created for the Physionet competition was further analyzed using SHAP. We found out that the model learned almost exactly the same diagnostic criteria used by cardiologists.

Multi-Label Cardiac Abnormality Classification from Electrocardiogram Using Deep Convolutional Neural NetworksThis is the paper published for the Physionet competition and presented in the CinC conference. We were able to create a model that classified 26 cardiac abnormalities from reduced-lead ECGs.

An AI Powered User friendly Neonatal Seizure Detection SystemWe designed an active-dry-electrode-based wireless EEG device with real-time monitoring, seizure detection, automatic artifact removal, and interpretability. We tested our device on children with absence seizures in realtime and device performed succesfully.

Electrooculography based Computer Interaction SystemEOG signals obtained from around the eye are analyzed to record the horizontal and vertical movement of the eye. The obtained signals are then used to classify the signals and control a computer game.

Smart stethoscope with ECGIntegrate a MEMS microphone in a custom diaphragm to get Phonocardiogram signals together with 3 lead ECG signals to create a smart stethoscope.

RISC-V Architecture based ProcessorImplement a processor on an Intel Altera DE2-115 FPGA board based on the RISC-V micro-architecture and instruction set. We were able to run a whack-a-mole game on the board.

Landfill Methane Monitoring and Analyzing systemA low-power IOT device to measure and analyze the methane levels found in landfills. The analyzed methane levels are transmitted over large distances through NBIOT*. The data is captured and stored in Microsoft Azure databases. Azure services are used to notify and alert users. Long term collected data will be used to predict methane levels using machine learning.

Reduced lead ECG classification using Deep CNNMachine learning model created using a Deep CNN to classify 26 cardiac abnormalities using reduced lead ECGs. The model is interpreted using Shapley additive explanations.

PID-based Hot Plate Temperature ControllerA hot plate temperature controller with a high output power is implemented using a PID controller to bring the temperature to a desired value.

Analog Implementation of a Simplex Line Coding TransceiverThis is our second lab project (group) done in semester 3. Here we mainly used the recently learned knowledge of analog electronic components. Our objective was to fully encode a bitstream in manchester, transmit over a large distance, and decode it in the receiving end; only using analog components.

Alarm ClockThis is our first lab project (group) done in semester 2. We were able to use our knowledge of PCB Designing, Enclosure Designing, AVR-C coding (microcontroller programming) in this project. We made an alarm clock from scratch.