Superhuman, the productivity platform that includes Grammarly, Coda, Superhuman Mail and Superhuman Go, serves over 40 million daily users across dozens of languages. As the company expanded from writing assistance into a multi-product AI platform, engineering teams found themselves spending more time maintaining custom sync infrastructure than building features. By adopting Lakebase and Databricks Apps, Superhuman replaced complex pipelines with a managed, transactional layer, cutting feature onboarding from months to weeks and freeing go-to-market teams to ship data-powered tools rather than waiting on engineering.
What is Lakebase?
Lakebase is a fully managed PostgreSQL database natively integrated into the Databricks Platform. It provides a transactional data layer that sits alongside Delta tables, enabling teams to sync lakehouse data into a low-latency relational store for application reads and persist application state back to Postgres for writes — without managing custom pipelines, alerting infrastructure or permissions separately.
What are Databricks Apps?
Databricks Apps is a deployment environment built into the Databricks Platform that allows engineering and data teams to build and host internal applications without configuring external cloud infrastructure, such as ECS instances or networking layers. Apps deploy within the Databricks security perimeter and integrate directly with lakehouse data, enabling you to move from prototype to production without a separate DevOps workstream.
The Infrastructure Challenge: Custom Pipelines and Manual Workflow
Superhuman runs one of the largest AI productivity platforms in the world, processing text across dozens of languages for millions of users every day. A unified data platform on Databricks powers analytics, machine learning and the pipelines behind customer-facing features. However, the infrastructure connecting that data to production applications and go-to-market workflows had grown fragile, creating a significant engineering burden on two fronts.
How a Redis sync failure exposed Superhuman's pipeline fragility
On the infrastructure side, the ML team had built a custom pipeline to sync data from the Databricks Lakehouse into Redis and DynamoDB. The pipeline powered eligibility rules for promotions and free trials, things like whether a user had activated a specific feature or logged enough sessions to qualify for a discount. "It was the right solution when it was built. But priorities shifted, the team moved on to other things, and the system stayed behind," said Michael Kobelev, ML Infra Software Engineer. Recently, a routine library update broke sync jobs writing to Redis. Without centralized alerting, the team couldn't identify all the affected jobs and some failures didn't surface until weeks later.
The last-mile gap between warehouse data and sales workflows
On the go-to-market side, the gap between data and action was just as costly. Sales and customer success reps manually copied metrics from dashboards into PowerPoint templates, spending roughly 30 minutes per deck, multiple times a week. "Data teams build incredible pipelines and tables, but there's still a last mile between what's in the warehouse and what a sales rep can actually use," said Maximilian Proano, Software Engineer, Data Applications. When the team built Deckster, an LLM-powered app that automatically generates customer-facing presentations, the first version pulled metrics live from a SQL warehouse. Latency spiked on every click, and there was no clean way to cache results or persist a user's progress.
Both teams needed the same thing. A managed transactional layer that sat close to their Delta tables and required minimal upkeep.
How Lakebase and Databricks Apps Replaced Custom Infrastructure at Superhuman
Lakebase eliminates sync pipeline maintenance for ML teams
Rather than reinvest in the legacy system, the ML Infrastructure team evaluated Lakebase, a fully managed PostgreSQL database natively integrated into the Databricks Platform. The automated sync pipelines were the decisive capability. Previously, product teams had to deploy and maintain their own pipelines, configure alerting and manage permissions. “Lakebase eliminated an entire layer of sync infrastructure that we used to manage ourselves," Michael explained. The team piloted Lakebase alongside Redis for the special offers use case and found comparable performance with significantly less operational overhead. Product engineers onboarded a new feature in under an hour by following a README and creating a merge request. Larger integrations have seen similar gains. A previous project that routed data from Databricks back into a production service took nearly three months on the old infrastructure; a comparable integration on Lakebase was completed in two weeks.
Databricks Apps standardizes internal application deployment
In parallel, the Data Applications team standardized on Databricks Apps to turn prototypes into deployed internal products without treating hosting and networking as separate projects. "Before Databricks Apps, every new application meant spinning up ECS instances, configuring networking, and solving the same infrastructure problems all over again," said Zhenwei Hu, Software Engineer on the Data Applications team. The team built reusable Claude Code skills and CI/CD workflows so that anyone, including data scientists and PMs, could go from idea to deployed app in about 30 minutes.
Lakebase as the transactional backend for bidirectional data flow
As those apps became more interactive, the team adopted Lakebase as the default transactional backend. "We needed data flowing in both directions. Delta tables sync into Lakebase for low-latency reads and visualizations, while user state, feedback and metrics write back to Postgres. Lakebase handled both, and the performance difference was obvious from day one," said Hubert Pham, Software Engineer on the Data Applications team. Data stays within the secure perimeter of the Databricks environment, and the broader organization, already familiar with the platform, requires minimal onboarding. "Databricks lets us focus our time on the problems that matter to the business, not on wiring infrastructure to solve them. That's what changed for us," said Maximilian.
Faster Feature Delivery and Lower Operational Burden
The shift to Lakebase and Databricks Apps has changed how Superhuman builds and delivers tools. What previously required quarterly release cycles now operates on a weekly cadence. The Deckster application alone saves approximately 7 hours per week for early adopters by automating tasks that previously required manual steps.
Teams are shipping faster across the board. Feature onboarding that previously took Superhuman's Data Apps team nearly three months on the old system now takes about two weeks, and simple data integrations can be completed in under an hour. Operational burden has dropped in parallel. On-call disruptions that once consumed three days per shift have fallen to roughly two hours as managed infrastructure replaced manual pipeline maintenance.
The team is already expanding on this foundation. Deal Analyzer will surface sales performance data from large tables, with Lakebase handling reverse ETL into a responsive application layer. On the infrastructure side, demand continues to grow to feed historical user data into production services for personalized prompts, improved language detection and targeted in-app messaging. "Lakebase has become the standard," said Hubert. "Any new application that needs ETL or persistent state, we start with Lakebase in the architecture from day one."