Hello! I am an NLP/ML enthusiast currently working as an NLP Research Engineer. My current work delves into supervised fine-tuning of LLMs, parameter efficient fine-tuning of LMs and LLMs, distributed training pipelines for these LLMs, RAG based use cases for LLMs. I am also exploring unsupervised topic classification models.

I completed Masters of Science degree in Intelligent Information Systems (LTI department) at Carnegie Mellon University, Pittsburgh. I have interned at Amazon Alexa AI as an applied scientist in the past, working on fairness in Entity Resolution models. In the past, I have worked with Prof. Alan W Black on downstream tasks like keyword tagging for low resource language speech. Currently, I work at a private investment research firm, as an NLP Engineering Expert. I work on parameter efficient fine-tuning of language models(LMs) like BERT useful for the finance domain. I am also working on distributed training of such LMs and taking care of the ML pipeline end-to-end.Between 2018-2021, I worked as a Software Development Engineer 2 at Goldman Sachs, Bangalore and contributed to some very interesting projects. I completed my BE Degree in Information Science and Engineering from PES Institute of Technology, Bangalore (Batch of 2014-2018).

I’m most interested in tackling problems in the Machine Learning, Natural Language Processing domains and building scalable, secure ML systems.

Publications

• Multi-Dimensional Evaluation of Text Summarization with In-Context Learning
  Sameer Jain, Vaishakh Keshava, Swarnashree Mysore Sathyendra, Patrick Fernandes, Pengfei Liu, Graham Neubig, Chunting Zhou
  Findings of the Association for Computational Linguistics: ACL 2023, Toronto, Canada

• Real-Time Headgear Detection in Videos Using Deep Learning Based Feature Extraction with A Supervised Classifier
  Swarnashree Mysore Sathyendra, Ranjana Seshadri, Rajdeep Pal, Dr. S Natarajan
  24th International Conference on Advanced Computing and Communications (ADCOM 2018), IIIT-Bangalore, 2018
  [Paper]

Updates

• 05/2023 Our paper on multi dimensional evaluation of text generation tasks using LLMs got accepted into ACL Findings 2023
• 31/08/2021 Started my degree at CMU
• 04/01/2021 Received an accelerated promotion to Software Developemnt Enginer 2)(Technical Asssociate) at Goldman Sachs
• 08/11/2019 Attending Grace Hopper Celebration India 2019 as a Speaker, representing Goldman Sachs

Select Projects

Please refer to CV for more details on all recent projects

• Multiparty Conversational Emotion Recognition:
  • Implemented an RNN-based approach to use \textbf{prior acoustic and emotion context} to predict future emotional state with performance improvements of 3% in weighted-F1 score over non-contexualized models
• Real-time Person Detection in Videos based on Natural Language Description
  • Built a working end-to-end system comprising of various modules that leveraged on deep learning models trained with fine-tuning.
  • Modules: Human Figure Detection, Body Segmentation, Gender Detection, Clothing and Apparel Style Detection.
  • This project led to a publication at ADCOM 2018.
• Factoid Based Question Answering System for Biology Coursebooks
  Built a Q&A system using QANUS skeletal framework. Devised and implemented custom chunking and parsing algorithms.

Professional Experience

• Smart FAQ Chatbot
  • Worked on building a smart FAQ chatbot that was trained on various FAQs in the Legal and Compliance domain.
  • Took up complete ownership of the entire lifecycle of the product that also included building an efficient solution for retrieving answers for the questions (Implemented word-jumbling techniques, auto correct using hunspell, SVD for dimensionality reduction and ConceptNet based semantic parser).
• Data Ingestion Pipeline
  • Currently building an ingestion pipeline and one-stop store for division-wide data, reducing consumption time for downstream applications from 48 hours to less than 1 hour.
  • Leveraged Presto as query engine and spark optimizations, yarn for resource management

CV

Scholar

LinkedIn

Mail: ms dot swarnashree at gmail dot com

last updated: June 16, 2024