Writer.com pens its gen AI success story with Google Cloud databases

Editor’s Note: As Writer, a full-stack generative AI platform, started to experience increased demand, its database requirements for capacity and capability grew dramatically. With AlloyDB, Bigtable, Datastream, Firestore, and Memorystore database services, Google Cloud provided security, flexibility, and ease of management, helping Writer to save on time and skilled resources — and to stay ahead of the technology curve.

Writer is a full-stack generative AI platform for enterprises. Founded in early 2020, we empower entire organizations with the ability to accelerate growth, increase productivity, and ensure compliance.

Writer transforms work by delivering high-quality outputs that are accurate, compliant, and on-brand. Our platform consists of Writer-built LLMs, a Knowledge Graph that connects our models to our customers’ internal data sources, AI guardrails to enforce our customers’ rules, a flexible application layer, and an ecosystem of robust APIs and integrations.

Leading enterprises across a wide range of industries choose Writer, including Vanguard, Intuit, L’Oreal, Accenture, Spotify, and Qualcomm, and the platform doesn’t use customer data to train models, while complying with SOC 2 Type II, HIPAA, PCI and GDPR.

How we got started

In our early days, we built out a full Palmyra large language model (LLM) training pipeline, enabled our software-as-a-service (SaaS) platform to use that model, and delivered use cases for specific business functions ranging from marketing to product and support.

Today, Writer is generating 90,000 words a second with over a trillion API calls per month. This, along with the recent generative AI market explosion, called for a major shift to our business. However, building an entirely new offering called for greater compute resources and automation. To make this possible and capitalize on the early move, we had to make sure our backend needs, including databases, were fully met. With the immense growth of our compute needs, managing MySQL would have required vast amounts of time and effort – even our first model started at 128 million parameters. An increased focus on security and the flexibility to accommodate growing customer requirements was another imperative. And all of this had to come with ease of management as we didn’t want to build out a large database and analytics team. We wanted to direct that time and effort to focus on our customers.

Google Cloud databases delivered, allowing us to re-architect and scale securely and without a massive increase in our database staff.

A family of databases that’s the “write” fit for us

Google Cloud managed database services enable us to store our customers’ data that they create and store while using Writer. To make this possible, our SaaS AI platform uses AlloyDB for PostgreSQL, Memorystore, and Bigtable as the supporting layer for approximately 100 microservices that we run on Google Kubernetes Engine (GKE). We’re also exploring AlloyDB AI’s vector query capabilities to power semantic search experiences.

Google Cloud databases offer built-in encryption, authentication and other security features; ease of scalability and cost control; and simple, tight integration with other cloud services such as GKE, BigQuery, and Datastream for BigQuery, which were lacking in our previous cloud service provider. This, coupled with Google’s intuitive user interface, ease of use, and quick pace of cloud platform feature releases, has been a better fit for us and matches our technology priorities.

We’ve always run a lean technical team focused on our core business requirements. Google Cloud databases allow us to manage a highly sophisticated and complex platform securely. Our ability to be highly effective with our technical talent and scale our output efficiently is an incredible competitive advantage. In combination with other Google services including the latest G2 Compute Engine GPUs, we have the speed and power to work extremely fast with real-time workloads and can handle heavy computation. Another benefit has been our ability to offer data residency to international customers who need their data to stay within their jurisdictional boundaries.

Even the migration from MySQL on VMs to AlloyDB for PostgreSQL was simple and painless. The success of this migration allowed us to dictate the pace of our future migrations, and the ability to choose which microservice to migrate, when meant that they were minimally impacted as we shifted them across databases. In addition, cost and performance management is now simpler — ‌we literally watched the usage shift in real time on our internal dashboard as we moved off of MySQL and onto AlloyDB.

Writing our own ticket with Google Cloud

We see Google Cloud databases supporting our future business goals too. The full text search capabilities of AlloyDB for PostgreSQL are helping us simplify and improve performance of our microservices collection. We also look forward to offering data residency as a regular platform feature, as well as the potential of reducing our costs further by optimizing our Memorystore and Bigtable usage.

Google continues to be our trusted partner, and Writer is thrilled to be leading the transformation and innovation of generative AI to enterprises together with Google Cloud.

Learn more

Posted in

Share this article
Shareable URL
Prev Post

Enterprise Connect 2024 – Bringing AI to the Contact Center

Next Post

The evolution of play: From live to living games

Leave a Reply

Your email address will not be published. Required fields are marked *

Read next