Selwyn Gomes

Working on XR software with a focus on improving system performance and latency under high workloads.

Team: XR software team. Working on the Snapdragon Spaces developer platform.

Provided support for the end-to-end deployment of machine learning models for low latency inference on Alveo FPGA boards leveraging the Xilinx Vitis flow. The contribution is made towards the hls4ml open source project.

• on providing theoretical guarantees and experimental evidence for using 1-bit compressive sensing with generative priors. • Explored a relaxed optimization problem in Gaussian process bandits and put forward theoretical bounds along with modified acquisition algorithms for faster convergence in the given setting. • Investigated the potential application of gradient-based methods in high dimensional Black-Box optimization using a probabilistic posterior function over the derivative space.

◦ Involved in developing Trading algorithms called alphas. ◦ The alphas are built using market indicators and ideas from Financial literature which are expressed with the help of statistical operators. ◦ The trading algorithms which are developed have high fitness and low turnover.These alphas are expressed on a web based platform called websim.

◦ In this work, we implemented a sequence to sequence model to construct a spectral graph from high dimensional input data by extracting the number of high-frequency events from the given sequence. ◦ We created a simple deep learning model using a combination of recurrent, convolutional and fully connected networks to perform the task using minimum external hardware.

ECE 271A - Statistical learning 1 ECE 225A - Prob Stats for Data Science CSE 258 - Recommender Systems and Data Mining PHY 245 - Parallel Computation ECE 143 - Python Programming for Data Science CSE 202 - Convex Optimization ECE 269 - Linear Algebra CSE 250A - Probabilistic Reasoning LIGN 167 - Deep learning and NLP ECE 277 - GPU Computing ECE 167 - Computer Vision