Nicholas Wright

Recent work:

• Redesigned poorly implemented API serving a suite of BERT-based text analytic models into a queue consumer pattern, resulting in a 90% reduction in operational costs
• Developed question answering and summarization tool utilizing OpenAI models supporting our professional consulting services to provide in depth text analytics
• Researched efficacy of different embedding models to support a Retrieval Augmented Generation (RAG) system (both proprietary and open source) and techniques aimed at minimizing embedding storage requirements; primarily dimensionality reduction and quantization
• Developed cutting-edge Natural Language Processing (NLP) system by fine-tuning open-source LLMs via LORA, focused on uncovering low level actionable insights within a piece of text rather than merely generating high level tags.
• Implemented API's and batch processing functionality for the NLP system using VLLM, Triton Inference Server, Redis caching, and Azure Functions
• Used creative spline regression technique, coupled with derivative analysis of the spline objects, to identify customer experience survey features with the highest expected ROI.
• Mapped survey features to low level text features to provide actionable suggestions to realize said ROI

• Designed and built customer facing suite of BI dashboards for over 100 clients, allowing customers to access and interact with SMG’s proprietary location tracking customer experience SurveyMini dataset
• Created R and Python scripts for running custom analyses of SurveyMini data not supported within BI dashboards
• Provided professional analytics services for clients, combining standard customer satisfaction data with ocation tracking data from SurveyMini
• Supported research analysts throughout the company in understanding how to use the BI dashboard to do ad-hoc analyses for their clients.

• Analyzed customer satisfaction data to provide clients with insights to improve their organizational performance.