Welcome to Ranak Homepage!
I’m a full-time Applied Scientist II (Level 5) with Amazon Web Services (AWS) AI Labs, specializing in forecasting and Large Language Models (LLMs) within the AWS Supply Chain group. I develop generative AI-powered interactive forecasting assistant that provides reliable and transparent answers to supply chain events - explain what is happening in the supply chain, why it is happening, recommend actions to take, and explore what-if scenarios to assess trade-offs and optimize supply chain decisions.
Prior to this, I was a Ph.D. Student at the University of California, San Diego, where I built efficient learning systems for label scarce sensory applications. I am grateful to Qualcomm and Halıcıoğlu Data Science Institute for their fellowship to support my research. During my Ph.D., I spent several wonderful summers working as an Applied Scientist II Intern at AWS AI, a Data Science Intern at Bell Labs, and a Software Development Engineer Intern at AWS Redshift. I was also an invited keynote speaker at KDD 2024.
Research Interests
I am broadly interested in building applications from time-series data, including:
- Supply Chain - Improve forecasting for better inventory management, sales prediction, and production planning for retail and manufacturing.
- Human Motion & Health Tracking - Using kinematics data from wearable sensors to understand human motion and physiological and biometric data to analyze health.
- Audio Analysis - Improve accented speech-related tasks and music chatbots.
To learn more about my work, please look into my projects and relevant publications.
News
[September 2024] UniMTS: Unified Pre-training for Motion Time Series has been accepted to NeurIPS ‘24 (First pre-trained model for human motion time series).
[September 2024] Started my first full-time job as an Applied Scientist II (Level 5) at AWS Supply Chain in Bellevue, WA.
[August 2024] Successfully defended my Ph.D. Thesis on Efficient Learning for Sensor Time Series in Label Scarce Applications.
[April 2024] First survey on Large Language Models for Time Series accepted to IJCAI ‘24.
[September 2023] Physics-Informed Data Denoising for Real-Life Sensing Systems has been accepted to SenSys ‘23.
[August 2023] Gave an invited talk on PrimeNet: Pre-Training for Irregular Multivariate Time Series at SIGKDD ‘23 Workshop on Machine Learning in Finance, Long Beach, CA, USA.
[June 2023] Returned to Amazon Web Services (AWS) AI as an Applied Scientist II Intern at the Speech Science Group. Worked on MusicLLM.
[February 2023] Towards Diverse and Coherent Augmentation for Time-Series Forecasting has been accepted to ICASSP ‘23.
[November 2022] PrimeNet: Pre-Training for Irregular Multivariate Time Series has been accepted to AAAI ‘23.
[August 2022] Won the Qualcomm Innovation Fellowship 2022 for our proposal on Robust Machine Learning for Mobile Sensing.
[June 2022] Joined Amazon Web Services (AWS) AI as an Applied Scientist II Intern at the Speech Science Group. Worked on Accent Robust Speech Recognition.
Work Experience
Applied Scientist II Sep 2024 - Present
Explainable and Context-guided Forecasting
Building a generative AI-powered interactive forecasting assistant that provides intelligent answers to supply chain questions, helps visualize complex scenarios, and explores the trade-offs between different decisions. This tool enables inventory managers and planners to understand supply chain events, their causes, and recommend actions while evaluating various "what-if" scenarios.
Applied Scientist II Intern Jun 2023 - Sep 2023
MusicLLM: Music Understanding with LLMs
Built an LLM-enabled music understanding framework to solve music tasks, like captioning, retrieval, tagging, and classification in a text-to-text fashion. The model takes music file and task prompts as inputs and generate text responses. Used audio features from Encodec and an instruction-tuned FLAN-T5 LLM. [Slide]
Research Fellow Oct 2022 - Sep 2023
Context Aware Learning for IoT
Worked on two projects: 1) PILOT: Integrated physics-based constraints to improve data denoising for Inertail Navigation, CO2 Monitoring, and Air Handling in HVAC Control System [Paper | Slide], 2) ZeroHAR: Used sensor environment knowledge and fine-grained biomechanical information from LLMs to improve Zero-Shot Wearable Human Action Recognition.
Applied Scientist II Intern Jun 2022 - Sep 2022
MARS: Multitask Pre-training for Accent Robust Speech Recognition
Built a pre-trained model for accent-robust speech representation that improves performance on several downstream tasks, like Speech Recognition by 20.4% and Speaker Verification by 6.3%, across 12 minority accents with few minutes of data. [Slide]
Data Science Intern Jun 2021 - Aug 2021
Automated Ticket Resolution from Semi-Structured Log Data
Developed an ML pipeline to automate ticket resolution by conducting data cleaning, preprocessing, and visualization on time-series semi-structured system-level log corpus, followed by statistical feature extraction and classification. [Slide]
Software Development Engineer Intern Jun 2020 - Sep 2020
Model Explainability from SQL Surface for Redshift ML
Built a SHAP-based Explainability tool for AWS Redshift ML, enabling users to write SQL queries to introspect model predictions. Improved execution speed by 2x and reduced memory footprint by 90% over existing workflow in AWS Sagemaker.
Preview | Press Release | Blog | Documentation | Slide
Research Assistant Jan 2019 - Aug 2019
Real-Time Principal Component Analysis
Innovated a data dimensionality reduction algorithm for streaming data, that uses a data-driven heuristic to learn a robust, noise-resilient eigenspace. Algorithm proven experimentally by spectral analysis and Bhattacharyya Distance.
Conference | Journal | Poster | Slide
Projects
UniMTS: Unified Pre-training for Motion Time Series
First pre-trained model for human motion time series that excels in zero-shot action recognition across a broad spectrum of wearable devices, regardless of body placement or orientation.
Large Language Models for Time Series: A Survey
Pioneered the first survey on how LLMs can solve time-series tasks through prompting, time-series quantization, alignment of time-series with text, using vision as bridge, and using LLMs as output tools.
ZeroHAR: Contextual Knowledge Augments Zero-Shot Wearable Human Activity Recognition
Innovated a Zero-Shot Wearable Human Activity Recognition framework that integrates contextual knowledge about activities from LLMs and sensory environment information to improve Zero-Shot F1 Score by 20%.
MLO-MAE: Downstream Task Guided Masking Learning in Masked Autoencoders Using Multi-Level Optimization
Developed a Multi-level Optimized Mask Autoencoder that improves image classification by integrating model pretraining with fine-tuning to allow fine-tuning performance guide an optimal masking strategy for pretraining.
PILOT: Physics-Informed Data Denoising for Real-Life Sensing Systems
Established a physics-informed data denoising scheme that integrates physics equations as constraints into training loss, ensuring the cleaned data complies with fundamental physical constraints, thereby improving data quality.
SHARE: Unleashing the Power of Shared Label Structures for Human Activity Recognition
Built an encoder-decoder architecture to model label name semantics that harnesses the shared sub-structure of label names to distill knowledge across classes, improving accuracy of human activity recognition tasks by 1.7%.
MusicLLM: Music Understanding with LLMs
Developed an LLM with music integration that generates text responses, including music genre, instruments used, mood, and theme, based on music files. Used Encodec audio features in conjunction with FLAN-T5 LLM.
STAug: Towards Diverse and Coherent Augmentation for Time-Series Forecasting
Crafted a time-series data augmentation strategy that incorporates both time- and frequency-domain knowledge through mix-up and Empirical Mode Decomposition, respectively, improving time-series forecasting by 10.7%.
PrimeNet: Pre-Training for Irregular Multivariate Time Series
Crafted a self-supervised pretraining method for irregular, asynchronous time-series through sampling density-aware contrastive learning and time-sensitive data reconstruction techniques, improving few-shot downstream performance.
TARNet: Task-Aware Reconstruction for Time-Series Transformer
Developed a self-supervised task-aware data reconstruction technique that uses end-task knowledge to customize the learned representation to boost end task performance, improving time-series classification by 2.7%.
MARS: Multitask Pre-training for Accent Robust Speech Recognition
Built a pre-trained model for accent-robust speech representation that improves performance on several downstream tasks, like Speech Recognition by 20.4% and Speaker Verification by 6.3%, across 12 minority accents with few minutes of data.
ESC-GAN: Extending Spatial Coverage of Physical Sensors
Devised a globally-attentive multiscale Super-Resolution GAN that uses irregularly spaced sparse data to generate high-resolution spatiotemporal data for regions with no physical sensors, reducing reconstruction loss by 3.7%.
UniTS: Short-Time Fourier Inspired Neural Networks for Sensory Time Series Classification
Innovated a Fourier-transform (FT) inspired weight initialization and learnable Short-Time FT layer redesign to integrate time and frequency domain information, improving F1 score for sensory time-series classification by 2.3%.
Automated Ticket Resolution from Semi-Structured Log Data
Developed an ML pipeline to automate ticket resolution by conducting data cleaning, preprocessing, and visualization on time-series semi-structured system-level log corpus, followed by statistical feature extraction and classification.
RIoT: Towards Robust Learning for Internet-of-Things
Built a comprehensive evaluation strategy and a model-agnostic training framework to facilitate robustness of machine learning models to missing feature sets in time-series data, achieving a 22% improvement in robustness.
Model Explainability for Redshift ML
Built a SHAP-based Explainability tool for AWS Redshift ML, enabling users to write SQL queries to introspect model predictions. Improved execution speed by 2x and reduced memory footprint by 90% over existing workflow in AWS Sagemaker.
Generating Rationale in Visual Question Answering
Developed a model that generates a rationale behind selecting a multiple choice answer in visual question answering. Trained a joint visual-linguistic model to build multimodal image- and text-representation and a GPT-2 model to generate text.
Real-Time Principal Component Analysis
Innovated a novel algorithm for dimensionality reduction on streaming data that learns a robust, noise-resilient eigenspace representation using a simple, fast heuristic, achieving Bhattacharyya Distance of over 0.90.
Extracting Real-Time Insights from Stock Market and Twitter
Built and deployed a web application, based on Real Time PCA, to generate real-time analytics on streaming data fetched from Twitter, Google Search Trends, Bangladeshi Online News Portals and Dhaka Stock Exchange.
Stock Price Forecasting from Historical Prices and News Trends
Devised a multi-modal model to forecast stock prices based on historical prices and latest news trends. Used an LSTM to model the stock prices, Word2Vec to represent news headlines and an MLP to concatenate the features.
Publications
UniMTS: Unified Pre-training for Motion Time Series
Xiyuan Zhang, Diyan Teng, Ranak Roy Chowdhury, Shuheng Li, Dezhi Hong, Rajesh Gupta, Jingbo Shang
Conference on Neural Information Processing Systems (NeurIPS), 2024.
Paper
Large Language Models for Time Series: A Survey
Xiyuan Zhang, Ranak Roy Chowdhury, Rajesh Gupta, Jingbo Shang
International Joint Conference on Artificial Intelligence (IJCAI), 2024, Survey Track.
Paper | Code
ZeroHAR: Contextual Knowledge Augments Zero-Shot Wearable Human Activity Recognition
Ranak Roy Chowdhury, Ritvik Kapila, Ameya Panse, Xiyuan Zhang, Diyan Teng, Rashmi Kulkarni, Dezhi Hong, Rajesh Gupta, Jingbo Shang
[Under Submission]
MLO-MAE: Downstream Task Guided Masking Learning in Masked Autoencoders Using Multi-Level Optimization
Han Guo, Ramtin Hosseini, Ruiyi Zhang, Sai Ashish Somayajula, Ranak Roy Chowdhury, Rajesh Gupta, Pengtao Xie
[Under Submission]
Paper | Code
MusicLLM: Music Understanding with LLMs
Ranak Roy Chowdhury, Rohit Paturi, Sundararajan Srinivasan
[Under Preparation]
PILOT: Physics-Informed Data Denoising for Real-Life Sensing Systems
Xiyuan Zhang, Xiaohan Fu, Diyan Teng, Chengyu Dong, Keerthivasan Vijayakumar, Jiayun Zhang, Ranak Roy Chowdhury, Junsheng Han, Dezhi Hong, Rashmi Kulkarni, Jingbo Shang, Rajesh Gupta
Conference on Embedded Networked Sensor Systems (SenSys), 2023.
Paper | Slide
PrimeNet: Pre-training for Irregular Multivariate Time-Series
Ranak Roy Chowdhury, Jiacheng Li, Xiyuan Zhang, Dezhi Hong, Rajesh Gupta, Jingbo Shang
AAAI Conference on Artificial Intelligence (AAAI), 2023. [Travel Grant]
Paper | Code | Poster | Slide
SHARE: Unleashing the Power of Shared Label Structures for Human Activity Recognition
Xiyuan Zhang, Ranak Roy Chowdhury, Dezhi Hong, Rajesh K. Gupta, Jingbo Shang
Conference on Information and Knowledge Management (CIKM), 2023.
Paper | Code | Slide | Package
MARS: Self-supervised Multi Task Pre-training for Accent Robust Speech Representation.
Ranak Roy Chowdhury, Anshu Bhatia, Haoqi Li, Srikanth Ronanki, Sundararajan Srinivasan
[Under Submission]
STAug: Towards Diverse and Coherent Augmentation for Time-Series Forecasting
Xiyuan Zhang, Ranak Roy Chowdhury, Jingbo Shang, Rajesh Gupta, Dezhi Hong
International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2023.
Paper | Code | Poster | Slide
TARNet: Task-Aware Reconstruction for Time-Series Transformer
Ranak Roy Chowdhury, Xiyuan Zhang, Jingbo Shang, Rajesh K. Gupta, Dezhi Hong
SIGKDD Conference On Knowledge Discovery and Data Mining (SIGKDD), 2022. [Travel Grant]
Paper | Code | Poster | Slide
ESC-GAN: Extending Spatial Coverage of Physical Sensors
Xiyuan Zhang, Ranak Roy Chowdhury, Jingbo Shang, Rajesh K. Gupta, Dezhi Hong
International Conference on Web Search and Data Mining (WSDM), 2022.
Paper | Code | Poster | Slide
UniTS: Short-Time Fourier Inspired Neural Networks for Sensory Time Series Classification
Shuheng Li, Ranak Roy Chowdhury, Jingbo Shang, Rajesh K. Gupta, Dezhi Hong
Conference on Embedded Networked Sensor Systems (SenSys), 2021.
Paper | Code | Slide
RIoT: Towards Robust Learning for Internet-of-Things
Ranak Roy Chowdhury, Dezhi Hong, Rajesh K. Gupta, Jingbo Shang
[Under Submission]
Paper
Real-Time Principal Component Analysis
Ranak Roy Chowdhury, Muhammad Abdullah Adnan, Rajesh K. Gupta
Transactions on Data Science (TDS), Volume 1, Issue 2.
Paper | Poster | Slide
Real-Time Principal Component Analysis
Ranak Roy Chowdhury, Muhammad Abdullah Adnan, Rajesh K. Gupta
International Conference on Data Engineering (ICDE), 2019.
Paper | Poster | Slide
Education
University of California San Diego
Ph.D. in Computer Science Sep 2019 - Aug 2024- Specialization: Data Mining
- Advisors: Professor Jingbo Shang, Professor Rajesh K. Gupta
- Thesis Proposal: Data-Efficient Learning for Sensory Time Series [Slide]
- Thesis Dissertation: Efficient Learning for Sensor Time Series in Label Scarce Applications [Dissertation | Slide]
University of California San Diego
M.S. in Computer Science Sep 2019 - Mar 2022- Specialization: Artificial Intelligence and Machine Learning
- Advisors: Professor Jingbo Shang, Professor Rajesh K. Gupta
- Thesis: Learning Across Irregular and Asynchronous Time Series [Dissertation | Slide]
Bangladesh University of Engineering and Technology
B.Sc. in Computer Science and Engineering Jul 2014 - Oct 2018- Major: Artificial Intelligence and Machine Learning
- Advisor: Muhammad Abdullah Adnan
- Thesis: Real Time Principal Component Analysis [Dissertation | Paper | Poster | Slide]
Honors & Awards
- Invited Keynote Speaker at SIGKDD 2023 Workshop on Machine Learning in Finance, Long Beach, CA, USA. [Workshop | Slide]
- Qualcomm Innovation Fellowship 2022 for our proposal on "Robust Machine Learning for Mobile Sensing". One of the 19 winners among 132 participants across North America. [News | Proposal]
- Recipient of Travel Award: AAAI 2023, SIGKDD 2022
- Halıcıoglu Data Science Institute Graduate Fellowship 2019 to fund investigations in data science, in recognition of my research accomplishments. One of the 10 winners among 3906 applicants. [News]
- Selected to participate at The Heidelberg Laureate Forum 2020. [Certificate]
- Selected to participate at The Cornell, Maryland, Max Planck Pre-doctoral Research School 2019, Saarbrucken, Germany. [Certificate]
- Research Fellowship awarded by Department of CSE, UCSD to support graduate studies.
- Dean's List and University Merit Scholarship awarded by Department of CSE, BUET, in recognition of excellent undergraduate academic performance.
Services
- Program Committee Member: SIGKDD 2024
- Teaching: TA for CSE 158 - Web Mining and Recommender Systems
- Mentored 2 UCSD CSE MS students on a research project.