Hey,

My name is Akash. I am a master's student studying Computational linguistics at the University of Rochester (Go Yellowjackets !!). I enjoy coding, cooking, traveling, and clicking photographs. This website is an attempt to document everything on my mind in one place. Here I share some of my thoughts on various topics, I present some of the photos I clicked and of course my portfolio and resume for potential employers.

"Memento Mori”

Blog

Photographs

Rushress library

Campus Starbucks

Library Lot

Pitsford dairy farm

https://open.spotify.com/embed/playlist/4woP5l6JGT7hbhqkqBIiVX?utm_source=generator

Resume

<aside> 👋 I build resource-optimized, interpretable ML models with my code. I have two years of experience as an ML engineer where I collaborated with multiple partner teams such as Business, Product Management, Strategy, and Business Management to deploy solutions into production.

</aside>

Contact

📧 [email protected]

🔗 linkedin.com/in/akashchaudhari

👾 https://github.com/akashchaudhari98

Experience

Graduate Research Assistant

University of Rochester, USA (May 2023 - Present)

Developed a neural CKY parser to parse combinatory categorical grammars. Developed an encoder model to convert CCG types into embeddings and a decoder model to convert them back. Trained multiple combinator controller models implementing CCG rules to combine two embedded types. Brought all these models together to form a neural CKY parser. The goal of this research was to introduce logical information into sentence embeddings.
My thesis explores the development of a model aimed at identifying & retrieving a set of documents that describe the “exact same event” as articulated in a user document. The model incorporates a ranking mechanism to evaluate and prioritize the retrieved documents based on the extent and granularity of the information they contain. This research aims to enhance the precision of information retrieval in the context of “event-centric” document analysis.

Machine learning Engineer

Hansa Cequity, Mumbai, India – (Aug 2021 - Aug 2022)

Objective - Design and implement ML pipeline for insights extraction over call center call recordings
Implemented preprocessing workflows for noise reduction, speaker diarization, and multilingual data handling. Ensured reliable datasets, improving downstream machine-learning outcomes.
Finetuned LSTM and transformer-based models (Roberta) on these datasets for tasks like entity recognition, and topic classification. Trained statistical ML models from the above results and deployed all the models together as scalable ML pipelines for insight extraction.
Leveraged Hugging Face transformers and AWS SageMaker for finetuning
Established robust monitoring systems with Amazon CloudWatch for performance tracking, anomaly detection, and retraining. Enabled efficient operations and ensured model reliability.

Machine learning Engineer

Zinc Technologies, Mumbai, India – (July 2020-August 2021)

Zinc technologies use ML, and NLP to help writers write better books. We provide analysis and recommendations to improve scripts to perform well in the market.
- Developed ****NLP models for book analytics. Trained a classification model to identify the genre of a script, & mapped it to books that had previously performed well in the market in a similar category.
- Used transformer models like BERT to analyze changes in the genre as the story progressed. Trained models to compare the two books & rate them based on fluidity, reading complexity, & genre absorption.

Machine learning Intern

Tata Consultancy Services, Mumbai, India – (Dec 2019-March 2020)

Developed machine reading comprehension and document Q&A model to answer questions on a document uploaded by the user.