Over the past nearly 9 years, I’ve had the privilege of working in the AI and Machine Learning space at HubSpot, witnessing firsthand how the technology has evolved and started a complete transformation of the tech sector.
As an anecdote, I remember working with a distilled version of GPT-2 in 2018 to add smart compose functionality to our support inbox. The excitement when the model returned a semi-coherent answer is embarrassing to reflect on when compared to today’s Large Language Models.
One of the great perks at HubSpot is accessing a personal HubSpot instance, something that I set up during my onboarding days in 2016 and have rarely interacted with since. Today, AI is an extraordinary tool that is going to reimagine how businesses and employees carry out their day-to-day tasks. I’m restarting this blog to share my experience leading the AI Agents team at HubSpot, to dogfood our product and share my own personal experience of growing from an individual contributor on a niche ML team to an engineering leader responsible for 30+ HubSpot engineers.
1. Dogfooding HubSpot’s Products
As an engineering leader, I believe in leading by example. HubSpot’s Breeze AI features offer incredible tools to help businesses harness the power of AI, and this blog will serve as my personal sandbox to explore them. By using our own technology to build and maintain this blog, I aim to provide honest feedback while showcasing what’s possible with AI-driven tools.
Today we kicked off an internal series of Demo Day’s where engineers can showcase our latest progress on AI Agents. These Demo Days serve a dual purpose, a chance for engineers to showcase their work but also an opportunity for the team to get early feedabck on the products we develop. Watching the product designer and FE engineers walk through the design mocks and UI, I realised how unfamiliar I am with our product.
2. Sharing My Experiences in AI
When I started at HubSpot, machine learning was a niche discipline—a tool for specific challenges that required highly specialized expertise. Fast-forward to today, and AI has become a cornerstone of how we build products, thanks in large part to the paradigm shift brought about by models like OpenAI’s ChatGPT. AI is no longer confined to the backend; it’s now part of the everyday experiences we deliver to our customers.
When I first moved to the US and got my first job as a Data Scientist, I remember reading a 2012 HBR article: Data Scientist is the Sexiest Job of the 21st Century and thinking I’ve landed in the right field. In retrospect, the article was slightly off the mark and it’s Research Scientists who hold the title.
This blog will reflect on that transformation—how we’ve adapted, what we’ve learned, and where we’re going. Whether it’s building scalable machine learning models or navigating the ethical considerations of AI, I’ll share my thoughts and lessons learned.
3. Discussing Agent Frameworks
At HubSpot, one of the most exciting projects I’m currently involved in is our work on Agent frameworks. These frameworks are the foundation for building intelligent, adaptable systems that can proactively assist users in achieving their goals. This work is by far the most exciting and challenging work that I’ve been a part of to date. There is no industry blueprint for Agents but the consensus is definitely that Agents are here to stay. HubSpot has a highly opinionated tech stack, in 2025 we’re looking to merge that with the latest advancements in open-source Agent Frameworks to deliver a remarkable Agent Building experience for our customers.
---
This blog isn’t just about sharing my perspective—it’s about starting a conversation. Whether you’re an AI practitioner, an engineering leader, or someone curious about how technology is reshaping the world, I hope you’ll jump into the comments and share your perspective. I’ll be writing these posts myself, and aim to provide weekly updates, but as our AI Platform matures I hope to delegate more and more of the writing to my HubSpot Agent employees.