Data Scientist
Data Scientist with NLP, Computer Vision, Machine Learning and Deep Learning expertise. I have experience working with cross-functional teams across the world.
Skills
- Programming Languages: Python, SQL, PySpark
- Artificial Intelligence: Machine Learning, Deep Learning, NLP, Computer Vision, A/B testing, Hypothesis Testing, Statistical Modelling
- Cloud: Azure, GCP, AWS
- Software Development: Git, Docker, Streamlit, Gradio
Experience
Data Scientist, Fractal Analytics
July 2021 - Present
- DT-LLM Research
- Designed and executed experiments to analyze language generation capabilities of large language models (LLMs) for non-English languages
- Designed and executed 10+ experiments comparing various LLMs and translation services across 18+ metrics
- Performed Hypothesis testing to validate the results and recommend best-performing design
- Presented detailed analysis and recommendations, enabling the client to make architectural decisions for non-English chatbots serving 1.5 million+ users.
- MBL Open AI
- Assessed the potential of LLMs for automating coding tasks in collaboration with a major international bank.
- Designed and developed Streamlit applications to demonstrate the capabilities of LLMs in conjunction with Bigquery and Alteryx data and APIs, enhancing automation.
- Executed POCs using OpenAI’s GPT-4, resulting in a remarkable efficiency improvement of approximately 10-100 times in custom code conversion, with an 83% success rate for complex workflows to Python, SQL. - Additionally, conducted a POC for complex text to BigQuery SQL conversion using GPT-4
- MG Bot
- Created Django backend APIs and RAG framework for a Chatbot based on a person’s persona
- Created advanced algorithms to parse 100+ word documents into topics for creating a vector database for the persona and utilized Langchain library to create a chatbot.
- Project Phos
- Designed and developed the Smart Verification module for a leading reinsurance firm for automatic verification of usage restrictions on data and ensuring compliance using Deep Learning Algorithms and cloud APIs
- Automated 70% of the verification process, improving efficiency and reducing manual workload
- Implemented web scraping and data wrangling techniques to curate high-quality training data for the deep learning model.
- Created, deployed and managed a deep learning model to enable intent classification of legal text documents, improving the module’s functionality and increasing automation by ~100%
- Created inference pipelines on Azure Databricks to monitor model performance on new data, ensuring consistent accuracy
- optimized the module code to decrease the execution time of different submodules by up to 5 times
- Leadership: Mentored interns on NLP and Computer Vision for creating POCs and submodules for the project
- Nominated for “Client Champion Award” for outstanding work
Education
- B.Tech, IIT (BHU), Varanasi
Projects
- Anomaly Detection: Anomaly detection of ECG signals using various ML and DL techniques like CNN Autoencoders
- A/B Testing: Analyzed data from A/B test of a website for revenue optimization and improving the conversion rate for the website.