Close

Yudhik Agrawal

Data Scientist @ GS | Researcher @ IIIT-H

RESEARCH RESUME SOFTWARE RESUME

About Me

Driven by a desire to contribute to meaningful endeavors, I am dedicated to being part of organizations with a clear mission, cohesive teams, and a strong sense of purpose.

Hello! I am Yudhik, a passionate software developer and experienced competitive programmer currently serving as an Associate within Goldman Sachs's Product Data team. In this role, I actively contribute to a platform responsible for delivering high-quality data services that empower and drive the firm's operations.

Throughout my academic journey, I consistently ranked among the top 3% of students in Computer Science at IIIT Hyderabad, both during my undergraduate and master's degree. My research work within the Cognitive Science Lab under the advisorship of Vinoo Alluri , focused on unraveling individual differences, such as personality traits, through the analysis of musical behavior . In addition, I made significant contributions in the research surrounding improvement of song recommendation systems, with top-tier publications in ISMIR, ECIR, and BigMM. Notably, my most significant accomplishment was the publication of a groundbreaking research paper in Scientific Reports, a prestigious journal published by Nature. My exceptional achievements in research and academics were recognized through the prestigious Huawei Scholarship of Excellence.

Previously, I developed my proficiency in Deep Leaning working as a Research Assitant at the Centre for Visual Information Technology (CVIT) with Dr. Avinash Sharma , on projects involving 3D human body reconstruction, registration, and the generation of temporally coherent sequences of human actions. Our research findings were successfully presented at renowned conferences including WACV and ICCV. I also worked in Robotics Research Center (RRC) with Dr. K. Madhava Krishna, on avoiding Drone Collisions by Path Planning after doing 3D reconstruction of the surrounding obstacles(eg. Humans) which need not be static.

Beyond my professional pursuits, I am deeply passionate about maintaining a healthy lifestyle, exploring the realm of finance, and engaging in complex problem-solving activities.

Research

Dissertation

My Music and I: Decoding Individual Differences via Musical Behaviour


June 2021: Master's Thesis in International Institute of Information Technology Hyderabad

Yudhik Agrawal

The work presented in this thesis sought to investigate the influence of individual differences in active and passive musical engagement. Overall, the results of this thesis corroborate ideas of embodied music cognition and the role of bodily movement in musical experience and preferences. Furthermore, the thesis highlights the role of individual differences in musical behaviour. Taken together, these differences and related musical preferences thereof manifest as both active music-induced movement and passive engagement on online music streaming platforms.

Publications

Decoding Individual differences and musical preference via music-induced movement


Febraury 2022: In Nature Scientific Reports (SciRep)

Yudhik Agrawal, Emily Carlson, Petri Toiviainen, Vinoo Alluri

We propose a Machine-Learning model that uses participants’ free dance movements to predict personality traits, music preferences, and cognitive styles and further finds associations between dance movements and traits.

Multimodal Fusion Based Attentive Networks for Sequential Music Recommendation


November 2021: In 6th IEEE International Conference on Multimedia Big Data (BIGMM 2021)

Kunal Vaswani, Yudhik Agrawal, Vinoo Alluri

We tackle the problem of sequential music recommendation by exploiting multiple modalities (including acoustic content, lyrics, and tags) by proposing a multimodal fusion based attentive networks.

Personality correlates of Preferred Emotions through Lyrics


July 2021: In 16th International Conference on Music Perception and Cognition (ICMPC-ESCOM)

Yudhik Agrawal, Vinoo Alluri

We aim to understand the relationship between personality traits and the emotions, as represented in the Valence-Arousal (VA) space, conveyed through lyrics of their preferred music.

Personality Traits and preference for Instrumental Music on online music streaming platforms


July 2021: In 16th International Conference on Music Perception and Cognition (ICMPC-ESCOM)

Ramaguru Guru Ravi Shanker, Yudhik Agrawal, Vinoo Alluri

We aim to link an individual's personality traits to the number of instrumental music tracks in their long-term listening history.

Transformer-based approach towards music emotion recognition from lyrics


March-April 2021: In 43rd European Conference On Information Retrieval (ECIR)

Yudhik Agrawal, Ramaguru Guru Ravi Shanker, Vinoo Alluri

We proposed a Transformer-based network architecture that, given the lyrics, outputs the classification of Emotion Quadrants, in addition to Valence and Arousal Hemispheres.

GlocalNet: Class-aware Long-term Human Motion Synthesis


January 2021: In IEEE Winter Conference on Applications of Computer Vision (WACV)

Neeraj Battan*, Yudhik Agrawal*, Veeravalli Saisooryarao, Aman Goel, Avinash Sharma

We proposed a two-stage activity generation pipeline to synthesize a long-term (> 6000 ms) human motion trajectory across a large variety of human activity classes.

Towards Multimodal MIR: Predicting individual differences from music-induced movement


October 2020: In 21st International Society for Music Information Retrieval Conference (ISMIR)

Yudhik Agrawal, Samyak Jain, Emily Carlson, Petri Toiviainen, Vinoo Alluri

We proposed a Machine-Learning model that predicts individual traits from music-induced movement patterns and further find associations between dance movements and traits.

HumanMeshNet: Polygonal Mesh Recovery of Humans


October 2019: In ICCV-Workshop (2nd Workshop on 3D Reconstruction in the Wild)

Abbhinav Venkat, Chaitanya Patel, Yudhik Agrawal, Avinash Sharma

We proposed a multi-branch multi-task HumanMeshNet network that simultaneously regress to the template mesh vertices as well as body joint locations from a single monocular image.

Featured In


Experience

Associate

Goldman Sachs, Bengaluru

Product Date Engineering : Platform responsible for delivering high quality data services to enable and power the firm

  • Designed firmwide financial instruments’ data model by collaborating with global business partners and technologists across several domains.
  • Enhanced ETL flow for financial data by replacing legacy codebase with scalable architecture thus increasing performance by ∼50% .
  • Conceptualised and implemented AI framework to generate static rules for the reconciliation of financial products, reducing manual efforts by 80% .
  • Transformed multiple complex legacy reconciliation processes to a generic python job thus improving code structure, performance and eliminating technical debts.
  • Mentored new hires by ramping them up with functional understanding of derivative products and backend codebase.

Alloy Streaming : Self-Service Hosted ETL platform to let users configure and monitor their data pipelines

  • Enabled new business by implementing relational sink (MSSQL) for ETL flow.
  • Developed ETL platform’s validation by allocating resources per user to ensure consistent availability by decreasing downtime significantly.
  • Set up Kafka servers to perform acceptance testing for the clients’ pipelines.

Research Assistant

Cognitive Science Lab, IIIT-H

Successfully defended my Master's thesis under Prof. Vinoo Alluri. The reseach work sought out to investigate individual differences associated with musical behaviour in the context of active music engagement (Music-Induced Movement) and passive engagement (Online Music Consumption).

Summer Analyst

Goldman Sachs, Bengaluru

Worked on Data Ingestion Enhancements in Data Pipeline Framework, and did further exploration which would benefit the project in future.

Research Assistant

Center for Visual Information Technology, IIIT-H

Worked with Prof. Avinash Sharma, on 3D-Human Body Reconstruction, Registration and generating emporally coherent Sequence of Human Action.

Research Student

Robotics Research Center, IIIT-H

Worked under Dr. K. Madhava Krishna, on avoiding Drone Collisions by Path Planning after doing 3D reconstruction of the surrounding obstacles(eg. Humans) which need not be static.

Teaching Assitant

IIIT Hyderabad

  • ICS101 - Computer Programming - Monsoon 2018
  • IEC239 - Digital Signal Analytics & Appl. - Spring 2019
  • CSE251 - Graphics - Monsoon 2019
  • CSE481 - Optimization Methods - Spring 2020

Programming Club Coordinator

IIIT Hyderabad

The work involves giving seminars from basic to very advanced topics.
It also involves hosting contests and alerting about various important events like ICPC, Google Codejam, Facebook HackerCup and also keep the students motivated for contests and routinely discussions.

Education

Internation Institute of Information Technology

July 2016 - June 2021

B.Tech. and Master of Science by Research in Computer Science

This program is about science - but it is not limited to the way science is traditionally practiced. This is also about computer science and IT - but it is not about creating IT professionals. It is about the science that is emerging in the interface of the information revolution and the challenges posed by some of the grave issues and problems in today's world. From newer ways of acquisition and management of multivariate data on complex natural systems to newer ways of deriving knowledge from the data using novel methods of data visualization and computational simulation, the world today needs a different breed of enabled scientists who are ready to navigate this interface. This program is for those who have the inner desire to take on these grand scientific challenges.

DAV Public School, Gurgaon

April 2014 - March 2016

Senior Secondary CBSE



Rotary Public School, Gurgaon

April 2012 - March 2014

Secondary CBSE

Achievements

Projects

Music Emotion Recognition via Lyrics

Developed a Transformer-based network architecture that, given the lyrics, outputs the classification of Emotion Quadrants, in addition to Valence and Arousal Hemispheres.

View Project

Vidalysis

Developed a Software-as-a-Service which can analyze/interpret the video, trimming relevant part of the video and can also search through video using image or text. This project has 4 pipelines dealing with Trimming the video based on the query, indexing video based on text and making notes of the video, subtitle, and sentimental-analysis online and Video Dubbing using GANS.
Presentation Link

View Project

Deep 3D Human Mesh GUI

Developed a Tk GUI toolkit which finds 3D mesh of a human body from a monocular RGB Image/Video using state-of-the-art Deep Learning network.

View Project

StackOverflow UserQuery

Developed a search bar on top of the StackOverflow API which provides more relevant thread results based on the search and also re-order the answers based on various NLP techniques like text-similarity(USE), statistical analysis and semantic analysis.

View Project

Heart-Rate-Monitoring[PPG]

Developed an algorithm to estimate heart rate from the periodicity of the PPG signal as Each cycle of the PPG signal corresponds to a cardiac cycle. Fitness tracking is gaining massive popularity due to the advent of wearable devices that can track your vital signs. Heart rate monitoring is one such feature in many devices such as smart-watches and wristbands.

View Project

Artificial Connect 4

An AI bot playing connect 4 using alpha-beta pruning and minimax algorithm. Works on 5 levels depending on the depth till which the AI bot predicts the moves. We gain go much depth by using Zobrist Hashing.
Everything is done in C++.

View Project

Linux Shell

Developed a Bash like terminal in C using Linux system calls which includes user-defined commands, piping and redirection, and signal-handling.

View Project

OpenGL Tunnel Rush

Tunnel Rush is a fast-paced dodging game inspired by Slope and Run 3. In this you are moving in a tunnel full of color at a blazing speed and need to avoid obstacles in different shapes. also player needs to be very agile and precise to see the position of the incoming obstacle and quickly shift your position away to avoid it. Stay focused and break your record!

View Project

Texture Synthesis & Image Quilting

Texture Synthesis: Given a finite sample of some texture, the goal is to synthesize other samples from that same texture. In this, finding the nature of the image and adjusting the parameters are handled by the algorithm dynamically.
Texture Transfer: Take the texture from one object and ‘paint’ it onto another object.

View Project

Tor-Hop Control

Hops while using Tor is very dynamic which causes the searches and other things to be affected by the location of the hop. This decreases the effectiveness of using Tor to some extent. We provided a solution by managing the second last hop to a specific location which greatly improves the efficiency of using Tor.

View Project

Twitter Sentimental Analysis using WordCloud and HeatMap

Unlike every other Twitter Sentimental Analysis, this web app provides much more easy to control features like evaluation WordCloud for the query and it also has the algorithm for making HeatMap. This app also does Semantic Analysis and classify into many categories making it a much more powerful and useful.

View Project

Skills

Contact

yudhik100@gmail.com

Yudhik Agrawal