Deep learning intern

• Jan 2023 - Present

Working on processing long documents in CLM-team. Working on processing long documents in CLM-team.

ML Trainee - ML Summer School 2022

• Jun 2022 - Jul 2022

Machine Learning Intern

• Apr 2022 - Jul 2022

Developed model service on AWS LAMBDA with backend support on fast api and celery for delayed queuing tasks and made the existing architecture scalable and fast by a factor of 2x. Model testing on big data of scale 1e9. Updated existing models for NLP tasks by reducing latency from 612ms/step to 19ms/step with increased accuracy and quantization of existing architectures(Distill-BERT, T5-base , BERT , XLM). Developed model service on AWS LAMBDA with backend support on fast api and celery for delayed queuing tasks and made the existing architecture scalable and fast by a factor of 2x. Model testing on big data of scale 1e9. Updated existing models for NLP tasks by reducing latency from 612ms/step to 19ms/step with increased accuracy and quantization of existing architectures(Distill-BERT, T5-base , BERT , XLM).

Research Intern

• Dec 2021 - Mar 2022

Created subscription code for mqtt client for multiple publishers and improved existing architecture code by 1.5X(3000 message/seconds frequency.) Connected it to the mainframe database for further processing in the machine learning model deployed on cloud for drift compensation up to a scale of 1 nanometer. Received data from sensors in the data logger then connected it to publisher code. Created subscription code for mqtt client for multiple publishers and improved existing architecture code by 1.5X(3000 message/seconds frequency.) Connected it to the mainframe database for further processing in the machine learning model deployed on cloud for drift compensation up to a scale of 1 nanometer. Received data from sensors in the data logger then connected it to publisher code.

Data Science Intern

• Aug 2021 - Nov 2021

Development of a web scraping service (Selenium & beautifulsoup4), NLP(NLTK,spacy) and deep learning model. For the final tokenization and classification used BERT after applying the clustering algorithms for finding alike sentences out of the more than 50000 sentences. Development of a web scraping service (Selenium & beautifulsoup4), NLP(NLTK,spacy) and deep learning model. For the final tokenization and classification used BERT after applying the clustering algorithms for finding alike sentences out of the more than 50000 sentences.