Amazon

Data Scientist

boosting sales and supply chain efficiency by 20%

AWS Glue and Python-based AutoML tools including H2O

Utilized AWS Glue and Python-based AutoML tools including H2O to enhance predictive analytics models, boosting sales and supply chain efficiency by 20%.

TensorFlow and Keras to develop deep learning models

Leveraged TensorFlow and Keras to develop deep learning models, achieving 70% accuracy in predicting traffic congestion.

providing actionable insights that guided continuous improvements to chatbot functionalities

Conducted thorough data analysis to track user interaction patterns, providing actionable insights that guided continuous improvements to chatbot functionalities.

leveraging Wikidata and perturbed questions from HotpotQA

to capture the long tail reasoning in language models under false assumptions

Created a specialized dataset of over 1500 questions leveraging Wikidata and perturbed questions from HotpotQA to capture the long tail reasoning in language models under false assumptions.

demonstrating their effectiveness in handling complex QA tasks

Designed evaluation metrics and achieved benchmark detection accuracies up to 67% with GPT-4 and 70% with Llama-2, demonstrating their effectiveness in handling complex QA tasks.

through Agile-driven rapid prototyping and testing

Improved ML solution reliability by 5% through Agile-driven rapid prototyping and testing.

ensuring continuous, timely updates for the prediction models

Integrated TomTom’s REST API to fetch and process live traffic data, resulting in a 15% reduction in data processing time and ensuring continuous, timely updates for the prediction models.

Langchain integrated with GPT-4 to design and implement a natural language processing system

improved the relevance and accuracy of automated customer responses

Utilized Langchain integrated with GPT-4 to design and implement a natural language processing system that improved the relevance and accuracy of automated customer responses.

using XGBoost and GBM, focusing on anomaly detection and efficiency metrics

Developed and tuned predictive models using XGBoost and GBM, focusing on anomaly detection and efficiency metrics across large datasets (up to 50k rows).