Accelerating client's AI/ML projects:

• Building custom LLMs

• Utilizing GPUs/TPUs/HPUs/IPUs

• Optimizing models for inference via distillation and ORT

• Developing responsible, ethical practices

• Using cloud infrastructure effectively

I am leading a course on using Transformers in Enterprises - specifically about pre-training large language models.

Developing NLP applications and capabilities to expedite medical voice-of-customer insight generation.

Proficient at implementing solutions using nltk, gensim, spaCy, transformers, PyTorch, Flask, Dash, Redis, Docker, and Databricks for tasks such as classification, NER, topic modeling, clustering, emotion detection, semantic search, and summarization of unstructured text.

• Built and deployed a web app that uses sentiment analysis (LSTMs and BERT) to measure political bias in the titles of news articles. (sentimentr.nmbroad.com)

• Achieved 50% accuracy using LSTM and 64% using BERT on hand-labeled data for a 3-label classification task.

• Created interactive web app with a title-analyzing RESTful API using Flask and hosting on AWS EC2 with web serving via nginx.

• The app makes scheduled calls to major news providers’ APIs to retrieve recent article titles, and stores them in an AWS RDS PostgreSQL database for analysis.