2024 in Review: Building, Iterating, and Raising the Bar

2023 was a fantastic year, full of shipping and personal growth. I had my first big wins while positioning myself as a more versatile Product Manager and it set the bar high for what was to come.

Starting 2024, my goal was clear: build on the momentum of the previous year and aim even higher—and this year delivered. From taking full ownership of the compute offering at Vercel to shipping in-function concurrency, a transformative leap for traditional serverless, the milestones have been amazing. I also led the development and delivery of Custom Environments, a long-awaited feature for enterprise customers, and continued evolving Hive, Vercel's compute platform. This year wasn’t just about meeting last year’s bar. It was about raising it even higher.

Previously on my career

My first year at Vercel, starting in March 2022, focused on improving the build and deploy pipeline, and how customers interacts with their deployments. This included achieving a 50% improvement in average build times and releasing key features such as the Logs tab. It was a rewarding year that combined my passion for Vercel’s mission with impactful contributions to its core infrastructure and product.

In 2023, I took on larger challenges by expanding my scope and exploring greenfield areas, launching products like Conformance, Code Owners, and the Azure DevOps Extension, while enhancing deployment workflows. One of the year’s highlights was a major infrastructure overhaul that boosted build performance and enabled hardware differentiation, making it a standout year of innovation and collaboration.

From builds to runtime: Owning the full compute stack

Taking ownership of Vercel’s entire compute offering was a natural progression after managing the builds infrastructure and leading the creation of our compute platform with Hive. Expanding the scope to include everything that happens at runtime—Vercel Functions, Edge Functions, and Middleware—aligned perfectly with the ongoing efforts to improve the developer experience and platform performance.

This expanded focus was all about turning long-developed innovations into the new defaults. Streaming by default, bytecode caching, and cold start prevention had been in beta or development for some time, and my priority became ensuring they were shipped as part of the core offering. These features are no longer optional—they’re now the standard, raising the bar on what a compute platform should deliver and redefining expectations for performance, reliability, and modern web workloads.

Building Hive: Vercel’s compute platform

The major infrastructure overhaul I’ve been working on since its inception in late 2023, now known as Hive, is Vercel’s low-level, untrusted, and ephemeral compute platform, designed to power our builds. From the start, I’ve been responsible for setting the vision and direction for Hive, organizing its rollout, and prioritizing the use cases and benefits that deliver the most immediate value. By focusing on clear deliverables that provide tangible results, we’ve ensured steady progress while maintaining iteration velocity and shipping cadence. You simply can’t build an entire compute platform and wait until it feels “perfect” to ship it, you’ll probably never ship it, you have to be smart on picking workloads, and migrating them to the new platform. Having real customers using the product as early as possible has been critical to validating and improving it.

While Hive has been quietly driving our builds infrastructure since last year, 2024 was the year it stepped into the spotlight with two blog posts: a deep dive into Hive’s architecture and how it connects with private networks. The benefits have been significant: hardware differentiation so customers can choose the right machine size, a 30% improvement in build times and a reduction in provisioning time for push builds, like those required for Secure Compute users, from 90 seconds to just 5. Hive is a product I continue to build and refine, laying the groundwork for even more innovation in the future.

Revolutionizing compute: Shipping Serverless Servers

This year, I had the opportunity to ship one of the most exciting innovations in compute: in-function concurrency. It’s a game changer for workloads, combining faster performance with significant cost savings. Inspired by the original spirit of Node.js—eliminating CPU idle time waiting for I/O—we’re bringing that philosophy back to the web world, enabling efficient and scalable compute.

To bring in-function concurrency to life, I drove educational materials, organized the beta stages, and engaged closely with customers to validate the savings and gather actionable feedback. I also pushed for a more pragmatic, experimentation-driven approach, encouraging the team to prioritize iterative shipping and real-world testing, including a public beta. By maintaining a tight feedback loop with early adopters, we ensured the feature met real-world needs while refining its performance.

The concept itself is simple but powerful: for example, when an instance is waiting for a response from an external API, a new invocation can reuse the same in-flight instance rather than spinning up a new one. Since compute costs are based on GB-hours, this approach effectively cuts costs while maintaining scale-to-zero efficiency. The results have been incredible, with customers seeing substantial savings and loving the impact. This is the future of compute, and I’m excited to keep building on its success.

Shipping Custom Environments: Addressing customers pain points

Custom Environments have been one of the most requested features, especially by enterprise customers, and shipping it this year marked a significant milestone. This feature provides flexibility for customers with complex workflows while ensuring zero disruption to those who don’t need it.

Bringing Custom Environments to life was a long and rigorous process. Enterprise customers often have unique needs, underscoring the importance of recognizing that “you are not the customer” and requiring a more thorough and structured product discovery phase. This involved dozens of interviews to deeply understand their workflows and pain points, followed by close feedback loops with early adopters to refine the solution. The requirements and insights gained from this process were essential in delivering the right first iteration of the feature. It also meant rethinking core aspects of how Vercel operates, starting last year with the ability to stage production deployments. The result? A feature that has been embraced with enthusiasm and love by our customers.

Looking ahead to 2025

As 2024 comes to a close, it’s clear that building great products takes time, and there are no shortcuts. The successes of this year—like in-function concurrency, Custom Environments, and the evolution of Hive—are the result of multi-year efforts, iterating on solid foundations and setting new ones for the future. Great outcomes don’t happen overnight; they’re built step by step, with consistent focus and a clear vision.

The seeds planted in past years are now beginning to flourish, and the work we’ve done this year has laid the groundwork for even more innovation. Consistency remains key—not just in what we ship today but in how we build toward what’s next. As we move into 2025, I’m excited to keep raising the bar, delivering impact, and turning new ideas into reality.