Saksham Bassi

I am a Research Assistant at CILVR lab at NYU studying cross lingual transfer of Large Language Models (LLMs). I am pursuing MS in Computer Science at Courant Institute of Mathematical Sciences where I am specializing in Machine Learning sub-fields.

I spent summer of 2022 with Amazon Care to build a payment failure notification system to avoid unsuccessful payments. Prior to my masters, I was Data Scientist II at Glance, InMobi where I worked in the Personalisation team to enhance exploration of social media feed using Representation Learning models. Prior to that, I spent time at HSBC as a Software Engineer, where I built monitoring architecture and improved backend of financial applications.

In my past life, I have worked on deploying neural networks for classifying variable stars at IUCAA and in various time-series at Tata Institute of Fundamental Research.

Open to work: I am looking for Software Engineering / Machine Learning roles starting June 2023.

Email  /  Resume  /  LinkedIn  /  Twitter  /  Github

Apart from having interests in software engineering, my research interests lie in natural language understanding, model generalization, and machine learning. I have had a great time applying machine learning models in interdisciplinary fields like astronomy, medical imaging, finance, industrial IoT.

One dimensional hybrid model of convolutional network and LSTM to classify variable star classes.

Machine learning technique for time series which combines statistical features and neural networks to model, factorize and reconstruct data.

Analyzing the structural patterns in time-series of Onion retail prices using various mathematical techniques to conclude about the patterns and periodicity.

Identifying diseases in farm fields using computer vision techniques on drone images. Transfer learning of pre-trained Convolutional Neural Networks and unsupervised segmentation as a pre-processing for training neural networks from scratch.

Convolutional Neural Networks to predict the kind of skin cancer using transfer learning and parallel networks

Source