859: BAML: The Programming Language for AI, with Vaibhav Gupta
Feb 4, 2025
auto_awesome
Vaibhav Gupta, Founder and CEO of Boundary, discusses the revolutionary BAML programming language designed to slash AI costs by up to 30%. He shares insights on natural language generation and how BAML streamlines AI interactions, enhancing data clarity. Gupta compares prompt engineering to traditional engineering, emphasizing its growing importance. Additionally, he reveals his unique hiring process, which prioritizes communication skills. Listeners will also learn about the use of retrieval-augmented generation technology and the future potential of BAML.
BAML, or Basic Ass Machine Learning, revolutionizes AI programming by introducing a more intuitive syntax that minimizes coding errors.
The development of BAML involved 13 pivots, illustrating the challenges and adaptability required in the startup environment of machine learning.
Boundary's unique hiring strategy prioritizes candidates' achievements and communication skills over traditional assessments, fostering a strong, innovative engineering team.
Deep dives
Introduction of BAML Programming Language
BAML, or Basic Ass Machine Learning, has been developed to streamline the interaction of developers with large language models (LLMs). By learning from the challenges of web development, BAML introduces a new programming syntax designed to be more intuitive and reduce common coding errors associated with prompts. This programming language significantly enhances productivity by improving clarity, allowing even project managers to utilize it effectively. It helps curtail possible mistakes in coding that could arise due to incorrect syntax, mirroring the way development tools like React transformed web coding.
Transformation Through Iteration
The journey of BAML's development included an impressive 13 pivots before the final product was realized, highlighting the challenges faced in startup culture. Originally aiming to compete with tools like Slack, the founders shifted their focus to addressing the burgeoning field of machine learning prompted by the rapid advancements seen with AI models like ChatGPT. This transition reflected their understanding that they had an edge in machine learning algorithms due to their professional history at tech giants and the insights gained through collaboration during their Y Combinator batch. The unique technical skill set of the founders became a defining factor in the creation of BAML, allowing them to focus on improving model interactions.
Efficiency in Prompt Testing
One of the standout features of BAML is its ability to improve the prompt testing process dramatically, allowing for quick iterations of model interactions. Developers can now run approximately 240 tests in just 20 minutes, compared to a mere five tests without BAML, owing to its hot reload feature. This rapid iteration is vital for developers working with AI as it significantly reduces friction, enabling them to experiment and refine their interactions without being bogged down by lengthy recompilation times. The efficiency of this process can lead to a more productive workflow, allowing engineers to focus on optimizing the functionality of their applications.
Addressing Token Efficiency
BAML enhances token efficiency through schema-aligned parsing, which minimizes the number of tokens required by intelligently handling outputs from models. This approach allows organizations to save 20 to 30% in operational costs by optimizing how data is structured and communicated to models, thereby reducing unnecessary tokens that would traditionally complicate interactions. The parser also operates seamlessly with existing models that may not support function calling, showcasing BAML's versatility and capability to adapt. The emphasis on token efficiency not only helps in cutting costs but also improves overall application performance and response times.
Innovative Hiring Practices
Boundary utilizes a distinctive hiring process that eliminates traditional job postings, instead inviting potential candidates to showcase their achievements through email by explaining why they are exceptional. This unconventional method prioritizes finding individuals who possess both technical skills and effective communication abilities, crucial for the work involved in developing BAML. Candidates undergo an in-depth reference interview instead of technical assessments, providing the company with deeper insights into their capabilities and collaboration skills. This approach fosters a team of engineers who share a passion for creating tools that make developers' lives easier, promoting a positive and innovative workspace.
In this week’s guest interview, Vaibhav Gupta talks to Jon Krohn about creating a programming language, BAML, that helps companies save up to 30% on their AI costs. He explains how he started tailoring BAML to facilitate natural language generation interactions with AI models, how BAML helps companies optimize their outputs, and he also lets listeners into Boundary’s hiring process.
This episode is brought to you by ODSC, the Open Data Science Conference. Interested in sponsoring a SuperDataScience Podcast episode? Email natalie@superdatascience.com for sponsorship information.
In this episode you will learn:
(04:53) What BAML stands for
(14:33) Making a prompt engineering a serious practice
(18:00) How BAML helps companies
(23:30) Using retrieval-augmented generation (RAG)