Shakiba Tasharrofi

Conducted research, developed, gathered data, and deployed a multi-stage recommender architecture integrating retrieval and ranking models using TensorFlow, enhancing channel admin visibility and convenience and improving user engagement. Conducted A/B testing demonstrating over 10% improvement in click-through rate (CTR) on text channels compared to the original app version without the recommender. Utilized Kubernetes for efficient and scalable deployment ensuring high availability and performance. Implemented MLOps practices using TensorFlow Extended (TFX) to monitor and address data drift, ensuring continued model accuracy. Developed and deployed a high-impact multi-output regression model using XGBoost and scikit-learn, achieving a significant 2.5% average monthly improvement in CTR, enhancing user engagement and ad effectiveness. Mentored two junior data scientists to build an effective churn prediction system, contributing to a remarkable 6% improvement in user retention. Develop

Productized the procurement categorization workflow into a reproducible, pipeline-based CLI tool using pure Python, enabling multi-level taxonomy classification with confidence scores and integrated feedback, significantly reducing operational time and enabling seamless cross-team collaboration. Designed, deployed, and scaled a Document Intelligence product with two production use cases (contract and invoice parsing) using GCP, Gemini, and LangGraph, processing over 30,000 invoices quarterly and resulting in a 3-year enterprise contract and successful customer adoption. Developed and deployed a Retrieval-Augmented Generation (RAG) and SQL-based chatbot identifying procurement opportunities, implemented with LangGraph, GCP, and Django, currently used by two companies to streamline procurement analysis. Contributed and developed a new web app section called Opportunity Dashboard leveraging Language Model Models (LLMs) for enhanced risk analysis, utilizing GCP services like Cloud Run for

Conducted research, developed, gathered data, and deployed a multi-stage recommender architecture integrating retrieval and ranking models using TensorFlow, enhancing channel admins' visibility and convenience and achieving over 10% improvement in click-through rate (CTR) on text channels through A/B testing. Utilized Kubernetes for efficient and scalable deployment ensuring high availability and performance. Implemented MLOps practices using TensorFlow Extended to monitor and address data drift, ensuring model accuracy continuity. Developed and deployed a high-impact multi-output regression model using XGBoost and scikit-learn, achieving a 2.5% average monthly CTR improvement. Mentored and guided two junior data scientists in building an effective churn prediction system contributing to a 6% improvement in user retention. Developed, deployed, and maintained a real-time personalized financial report generator for over 200,000 users, reducing report creation time from 60 seconds to 4 se