Internships
Algo Developer Intern
Hudson River Trading, New York
Developed a high-frequency trading execution engine for the Brazilian Equities; traded live on the B3 exchange.
Devised a market-making strategy for the firm’s Client Market Making Platform.
ML Intern
Hakimo, San Francisco
Designed a door detection algorithm for Hakimo’s security systems.
Developed a Panoptic segmentation model and a convex-hull-based polygon-approximation algorithm to obtain bounding boxes.
Created an algorithm to detect the secure side of a door for classifying applicable alarms. Reduced manual annotations by 92.45%.
Quantitative Trading Intern
Jane Street, Hong Kong
Researched on modelling equities to predict the next day opening prices of international stocks leading to a profitable trading strategy
Worked on the options desk to model future change in implied volatilities for various indices in Hong Kong, South Korea and India
Participated in multiple mock manual/electronic trading simulations which applied various trading concepts
ESTIMATION OF ALZHEIMER USING DEEP LEARNING TECHNIQUES
under guidance of Juned Kadiwala, University of Cambridge
Developed a methodology to pre-process the data
Worked on predicting Alzheimer by using audio and transcripts
Worked on various deep learning methods for acoustic as well as linguistic embeddings
Achieved high accuracy on manual and ASR transcripts
Coordinated team of 2 additional members
The ArXiv preprint can be found here.
The paper was accepted at BIOKDD 2021
IMAGE SEGMENTATION AND CLUSTERING
under guidance of Abhishek Hingne, PivotChain
Worked on a CO-attention Siamese Network (COSNet) to address the unsupervised video object segmentation task.
Designed and implemented an image-clustering algorithm based on similarity
It works by extracting features from a VGG model and applying clustering (K Means and DB Scan)
A demo for the clustering and a sample result for the unsupervised object segmentation algorithm can be found here
PERCEPTION MODULE FOR AUTONOMOUS VEHICLE
under guidance of Ujjwala Karle, ARAI
The algorithm is used to determine a depth of an object to take right navigation decisions by autonomous vehicle
The algorithm is implemented in python and works with correlation of camera image and Lidar image
ADDITIVE COMBINATORICS (READING PROJECT)
under guidance of Dr. Partha Mukhopadhyay, CMI
Additive combinatorics interfaces with combinatorics, number theory, ergodic theory, harmonic analysis and geometry over finite fields
Covered a major portion of Guth’s book “Polynomial Methods in Combinatorics”