Podcast

Ep. 16 – Andrew Hoh, Last Mile AI: Build vs. Buy is not a permanent decision

Author headshot
By Andrew Hoh
Back to all articles
Post hero image

## The Context: A Changing Tech Landscape

Andrew shares how his accidental entry into product management shaped his career, highlighting the spectrum of challenges faced when transforming machine learning (ML) research into tangible products. Andrew's journey set the stage for the discussion on how Last Mile AI is tackling the last mile problem in AI by enabling companies to test and evaluate large language models (LLMs).

## Deep Dive: Why Labeling Matters

A significant portion of the conversation centered around labeling—a critical element in honing the accuracy of AI models. For the uninitiated, labeling is the process of annotating data with tags used to train ML models. Andrew explained its importance in creating reliable AI systems and how Last Mile AI's approach helps in evaluating LLMs, ensuring that models are production-ready and robust against unpredictable behavior.

## The Core Dilemma: Build or Buy?

Andrew turned the conversation towards a pivotal build vs. buy decision during his tenure at Airbnb. At the heart of any successful AI deployment lies the tooling and infrastructure. Airbnb faced a crucial decision: should they develop a custom labeling tool or purchase an existing solution? Andrew discussed the intricacies involved in making such decisions, emphasizing the need for a thorough cost-benefit analysis and a deep understanding of immediate and long-term needs.

## A Noteworthy Solution: The Hybrid Approach

One of the key takeaways from the episode was the notion of a hybrid build-buy approach. According to Andrew, tech companies, especially those in their growth phases, often find themselves in a position where combining bespoke development with third-party solutions yields the best results. At Airbnb, this approach enabled them to leverage AWS’s ground truth for back-end operations while crafting a custom front-end tailored to their specific needs.

## Reflections on AI Development Andrew shared his excitement about the democratization of software development through AI, predicting a future where more individuals engage in building custom solutions. The accessibility of AI as a tool for fledgling developers promises a new era of personalized applications.

## Recap: Embrace, Evaluate, and Evolve

The episode wrapped up with reflections on the continuous evolution of build versus buy decisions, underscoring their nature as ongoing, rather than static, evaluations. With the tech landscape rapidly evolving, the capacity to adapt and mold decisions over time is a skill that can set companies apart. Whether you're a tech leader grappling with resource allocation or an AI enthusiast curious about the nuances of product development, this podcast episode is brimming with insight. It reinforces the importance of strategic foresight in navigating the complexities of modern technology investments. For a deeper dive into this intricate topic, tune into the full episode.

Get a demo
(from a real person)

Schedule some time on our calendar to talk through your specific use case and see which Anvil products can help.
    Want to try Anvil first?
    Want to try Anvil first?