Jun 2024 – Jul 2025 · 1 yr 1 mo · Bengaluru, Karnataka, India · Hybrid

• Worked with Product match and Pricing team,
• Specifically on exact looking products matching and grouping.

Apr 2024 – Jun 2024 · 2 mos · New York, United States · Remote

• Making devin better at software development tasks!

Jul 2022 – Jun 2024 · 1 yr 11 mos · Bengaluru, Karnataka, India · Hybrid

• Worked on creating docker service for annotating data for NER, and text classification tasks, with the help of open source LLMs like UniversalNER-7B, Llama2 7B and 13B with multi agent debate prompting mechanism.
• Deployed pretrained models like Flan-T5 xxl, Flan-ul2 for summarization and NER use case with AWS Sagemaker inference endpoint utilizing DJL for efficient model serving.
• Reduced the latency of a T5 based Setfit model by exporting them to ONNX and performing graph
• optimizations.
• Worked on Multilingual PostASR text correction task for reducing the inference of model from 600ms to 400ms per request, with 4bit float quantization.
• Worked on fine-tuning indicBART model (245M params) on multilingual data (hindi, tamil) and compared PEFT based lora finetuning and regular fine-tuning approach.
• Worked on NER for low-resource (tagalog-english) code-mixed languages.
• Developed model agnostic inference flask API with locust support for load testing the ML models. (supported model frameworks - pytorch, spacy, onnx)
• (Supported tasks - NER, classification)

Jan 2022 – Jul 2022 · 6 mos · Bengaluru, Karnataka, India

Jan 2022 – Jun 2022 · 5 mos · Remote

• Worked with naas.ai as an open-source contributor.
• > Contributed in solving issues, reviewing code, creating new issues linked to the creation of jupyter notebooks with Github API integration and creating low code drivers for the same.
• Find my work here -> https://github.com/jupyter-naas/awesome-notebooks/tree/master/Github
• > Worked on developing automation scripts for bringing the best of remote job opportunities in the field of data science direct to the slack channel of naas.ai (My first data product!)

Dec 2021 – Dec 2021 · 0 mo · Remote

Sep 2021 – Dec 2021 · 3 mos · Noida, Uttar Pradesh, India

• Joined as an first ever intern in the NLP division, worked under the founding engineers in contributing towards the product.
• Researched on implementing various experimentations (including dependency parser, canonicalization, etc) on fine-tuning BART base architecture for math word problems (used AWS EC2 instance for training the language models, and S3 for storing the model artifacts) and increased the accuracy to around 93%
• Contributed in developing scripts for augmentation (using Language models), canonicalization of numbers in word problems, and data processing.

Jun 2021 – Sep 2021 · 3 mos · Hyderabad, Telangana, India

May 2021 – Jul 2021 · 2 mos · Hyderabad, Telangana, India

• Worked on SentenceBERT for getting the sentence embeddings and using it for building a semantic content based recommendation engine. Also worked on keyword extraction using BERT. Deployed a prototype via Flask API on Heroku as well.
• Tech Stack: Keyword extraction with BERT, BERT Sentence Embeddings, Web scraping, OCR, Recommendation systems, Git, Heroku, Flask.

Mar 2021 – May 2021 · 2 mos · Bangalore, Karnataka, India

• Developed an algorithm for converting notes/text/chats (which contain various emojis, emoticons, etc) into some meaningful text, especially for users who need assistive technology using sentiment analysis.
• There are more than 1Billion people worldwide who are in need of assistive technology (like blind, visually impaired users) right now.
• Our solution ended up in the TOP-2 teams across participants coming from 130+ teams PAN India