Currently, I am a Senior Research Scientist at Buildnetic where I focus on designing and implementing automated data pipelines and applications that utilize advanced AI techniques for extracting and organizing unstructured data. I have also developed solutions for continuous streaming automatic speech recognition (ASR) for Indic languages, leveraging both in-house and Azure-based models, along with integration of denoising systems for enhanced speech recognition quality.
Previously, I was a Senior Research Scientist at Saarthi AI, where I worked on leveraging Generative AI for automation tasks, including developing a closed QA voice/chatbot system and optimizing models for chat-based applications. My work involved fine-tuning LLMs like Flan-T5 and LLaMA-2 and transitioning from intent-based to intent-less chatbot systems.
Before that, I was a Senior Research Scientist at Salesken AI, where I worked on a range of NLP downstream tasks, including text-to-text translation, semantic similarity, emotion recognition, sentiment analysis, and text summarization. I developed semantic similarity models using Sentence Transformers and built sequence-to-sequence models for understanding conversation context and predicting spoken dialogues. Additionally, I created customized versions of large deep learning models by optimizing network architectures and deploying them using tools like ONNX and TensorRT.
Earlier in my career, I was a Research Fellow at Microsoft Research India, where I was advised by Dr. Sunayana Sitaram. At Microsoft, I worked on developing Automatic Speech Recognition (ASR) systems for multilingual and code-switched conversational speech, employing techniques such as adversarial training and regularization to improve model robustness and reduce bias.
Earlier, I was a Research Intern at the Indian Institute of Science, Bangalore, India (IISc, Bangalore) under the supervision of Prof. Partha Pratim Talukdar. Here, I explored the use of knowledge bases to solve the problem of visual question answering.
I graduated with a B.Tech in Computer Science and an MS in Computational Natural Science from IIIT Hyderabad in 2019, where I was advised by Prof. Nita Parekh. For more details, check my CV or reach out to me via email.
Publications
Curriculum Learning for Adapting Models on Code-Mixed Data in Emotion Recognition
Dheeraj Agrawal, , Neeru Dubey, Shubham Sharma, Bharatram Natarajan, Suvro Banerjee and Ashish Kumar
Under Review | Coling'22
pdf
abstract
cite
video
Learning not to Discriminate: Task Agnostic Learning for Improving Monolingual and Code-switched Speech Recognition
Gurunath Reddy, , Basil Abraham, Vikas Joshi, Sunayana Sitaram
WSTCSMC 2020 |
First Workshop on Speech Technologies for Code-switching in Multilingual Communities 2020
pdf
abstract
cite
video
CoSSAT: Code-Switched Speech Annotation Tool
, Pratik Joshi, Sebastin Santy, Sunayana Sitaram
AnnoNLP@EMNLP'19 |
Empirical
Methods in Natural Language
Processing
pdf
abstract
cite
slides
Unsung Challenges of Building and Deploying Language Technologies for Low
Resource Language Communities
Pratik Joshi, Christain Barnes, Sebastin Santy, Simran Khanuja,
, Anirudh
Srinivasan, Satwik Bhattamishra, Sunayana Sitaram, Monojit Choudhury and Kalika Bali
ICON'19 | International Conference on Natural
Language Processing
pdf
abstract
cite
First Workshop on Speech Processing for Code-switching in Multilingual Communities: Shared Task on Code-switched Spoken Language Identification
, Sunayana Sitaram, Rupeshkumar Mehta
WSTCSMC 2020 |
First Workshop on Speech Technologies for Code-switching in Multilingual Communities 2020
pdf
abstract
cite