Associate Machine Learning Engineer at mavQ

• Designed and deployed Natural Language Processing (NLP) algorithms to extract key-value pairs from unstructured data, enhancing extraction accuracy by 90% using Gemini. 

• Utilized machine learning models and statistical techniques to enhance the accuracy and efficiency of key-value pair extraction by 90%. This enhancement was pivotal in successful deployments resulting in revenue of approximately 1M dollars. 

• Integrated YOLOv5 model for checkbox detection, improving document classification models, and reducing manual processing by 50%. 

• Built a document similarity model leveraging advanced NLP algorithms, achieving a 90% accuracy rate in real-world testing scenarios. 

• Proficient in implementing various deep learning architectures like Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) for effective preprocessing of textual data, including tokenization, stemming, lemmatization, and handling imbalanced datasets. 

• Engineered a machine learning model that classified document quality with a 92% accuracy rate, contributing to a project team’s success. 

• Led a team to implement SonarQube, enhancing the quality of all services, and improving maintainability by 25%. 

• Developed an efficient Text to SQL/SOQL conversion system using GPT models and LLMs, automating the process for structured data extraction with 80% of data extracted accurately with PII data masking.  

• Conducted research on prompt engineering techniques, enabling 80% structured data extraction from documents using LLMs, multimodal vertex AI, Lamaindex and Langchain, driving significant process improvements.