I have over 7 years of experience processing and analyzing text for hundreds of use-cases ranging from e-commerce to public health, and have published methods that combine data science with cognitive psychology. I can work with all traditional NLP tasks, and can build bespoke solutions for your unique problem space.
Services: Recommenders, classification, clustering, chatbots, knowledge graphs, sentiment analysis, product review analysis, topic extraction, user profiling, customer segmentation, psycholinguistics, and text processing.
Past Work: Using over 100 million Amazon product pages, I built an end-to-end, serverless solution into the GeniusWire application on AWS that classifies products with thousands of labels. The solution achieved 90% accuracy as well as being scalable, fault-tolerant, and cost-effective.
I have worked with pre-trained transformer technologies for over 3 years, and can build your LLM solution using recent advances in the space such as RAGs and multi-agent workflows. I can then deploy these applications to AWS for you for public or internal use.
Services: RAGs, multi-agent routing & workflows, LoRA/QLoRA PEFT, performance evaluation, prompt engineering, reranking, semantic similarity, graph extraction.
Past Work: I built an internal dashboard application in Streamlit that enabled my client's workforce to extract psychological insights about their online consulting engagements from transcript data. The application included a dashboard that visualized psycholinguistic features throughout a submitted transcript. It also included an internal RAG chatbot to engage with the transcript and all previous engagements with the client.
I have built deep learning solutions into several projects for my clients. Deep learning models come with a lot of technical debt so I am mindful to design solutions based on my client's current technology competency and needs. All delivered models include a report on model performance and provide clear documentation on how to work with the technology stack.
Services: Recurrent Neural Networks, Convolutional Neural Networks, Gated Recurrent Units, Auto-Encoders, Long Short-Term Memory.
Past Work: I built a deep-learning recommendation system for an early stage startup using embeddings and a GRU classifier. The model used cosine distance to recommend similar products based on their descriptions and current social media performance trends.