Love, UWS Thoughts

I'm Kevin Kirby, a Senior Product Manager in the Ad Platforms division of the Walt Disney Company (clears throat: conflict of interest statement), where I work on addressable ads data platforms, products, and infrastructure for streaming properties. I live on the Upper West Side of Manhattan in New York City with my fiancé, Mao-Lin, and serve as the elected treasurer of the board for the self-managed co-op we live in together, a position I was re-elected to with 90% of the vote. In my free time I'm an avid road cyclist and house and techno music junkie.

In summer 2025, I'll graduate from the Master of Data Science Program at the CUNY School of Professional Studies, where I recently finished my masters thesis on integrating large language models into enterprise system architectures. The research was grounded in my deep understanding of enterprise data and system architectures and the challenges and opportunities involved. As part of preparation and execution of the research, I built integrations with local LLMs and created a mechanism to bring structured data from a SQL database together with unstructured data embedded in a vector database.

Having spent the last year and a half teaching myself about machine learning and how models are built, deployed, and use, I now have some key AI development principles:

My Key AI Development Principles

• Rapid prototyping in secure sandbox environments that mirror production grade setups
• Taking the time (usually 80+% of a project) to clean and preprocess data so it’s suitable for meaningful use
• Establishing robust monitoring and evaluation from the outset
• Empowering teams by to iterate quickly and learn fast. Everyone can make a prototype.

All of this work means I now have direct and meaningful development experience with the following AI platforms and services:

• Hugging Face local and inference API LLMs
• Llama Index and Haystack orchestration frameworks
• Zero shot, single shot, and few shot prompting
• Phoenix Arize for observability
• DeepEval, Phoenix Arize, and RAGAS for measurement and evaluation
• Modal, Run Pod, Google Cloud Platform, and AWS for cloud computing
• Weaviate for vector and document storage
• Supabase, PostgreSQL, Databricks, Athena, and Snowflake for relational data

My next step is to drive innovation and responsible evolution the intersection of product management and machine learning. I want to help organizations drive implementation and adoption while not destroying their existing business.

Questions? Slide into my DM's via the Contact tab or email me directly at hello@uwsthoughts.com.