Hi, my name is
Nagraj
Data Scientist 💻  Â
I am currently seeking full-time opportunities and will be available to begin new professional challenges in the US from September 3, 2024, under F1 OPT visa status
About Me
Hey there! I am a Data Scientist based in the United States. Currently a graduate student at Purdue University.
I crunch numbers, unleash creativity, tell stories and build products. I have a special knack for understanding the business context behind products & services. A strong foundation in statistics and programming, including experience with databases and systems, has consistently fueled my success in delivering enterprise-level insights that drive growth. Now, I'm eager to leverage these skills and enthusiasm for AI and Deep Learning as an AI/ML Engineer, tackling complex problems and contributing to cutting-edge solutions.
If you're looking for an analytical mind with a passion for AI and a collaborative spirit, you've found him ! Let's connect!
I have experience with:
➣ Languages:
Python3, JavaScript, C, C++, SQL
➣ Tools:
AWS, Azure, Docker, Webpack
➣ Frameworks:
ReactJS, NextJS, Pytest, Pytorch
➣ Theory:
Foundational Models, Computer Vision, Optimisation Algorithms, System Design, Databases
Work Experience
Spry
SocGen
RAMAN
Spry Therapeutics Inc.
Data Scientist
February 2022 - March 2023Optimised pose detection models used in client-side AI, improving speed by 200% and accuracy by 65%
Proposed and led movement categorization project helping product scale from 5 exercises to 400+ exercises
Built novel algorithms for calculating 15+ accuracy parameters of 5 categories of exercises
Directed a 4 member data science team on Remote therapeutic monitoring (RTM) tool securing a $7M funding from marquee investors
Improved assessment reports by introducing visualizations using opencv and pillow
Developed LightGBM pose classifier reducing exercise identification false positive rate from 9% to 0.1%
Implemented Computer vision pipeline utilizing celery, fastapi, AWS cloud, and mediapipe
Projects
Multimodal Meme Search Engine
Developed a search engine using CLIP model for semantic search on meme images and text queries.
LLM Powered Autonomous Marketing Agent
Developed an LLM-powered agent to identify trends and generate platform-specific post ideas, boosting engagement by 25%.
Walmart M5 Sales Forecasting
Forecasted 30K+ time series using sales, pricing, and temporal features, achieving a validation loss of 0.013.
Decoding the 'Super' in AirBnb's Superhosts
Decoding the superhost status and its criteria over 16 evaluation periods. Understanding if it helps, how it helps and who it helps.
LLMs for Automated Medical Documentation
Hosted Llama-2 on AWS Sagemaker for zero-shot prompting and RAG, achieving WER of 0.86 on medical transcripts.
Data Driven Marketing for GAP Inc.
Exploring data driven marketing techniques for fashion retail industry including market segmentation, customer sentiment analysis, financial data analysis and SERP analysis.
Driver Performance Analysis Across F1 Circuits
Maximizing the excitement in a Formula 1 Grand Prix with some network analysis and algorithmic magic
Analysing clique behavior in academic settings
Understanding relationships among a real-world collaboration network of researchers who have published General Relativity work in Arxiv
Key Opinion Leader Identification for Neurology
Developed a data-driven approach using ML and NLP to identify and engage influential KOLs in neurology, enhancing market impact.
Retail Theft Analysis
Deep dive into analysing the surge in Retail theft for the cities of Chicago and San Francisco
The Fire Guide