Leave your details
Precisify
AI modules for building brand-safe YouTube campaigns
An ad operations specialist at Precisify needs to build a campaign: find YouTube channels that fit a product and its audience in a given market, and make sure none of them sit next to content a brand can't be near. Doing that by hand across YouTube's scale is slow. We build the tools inside Precisify's platform that turn it into a few steps, most of them driven by the company's own AI models.
The Client
Precisify is a media-buying platform for brand-safe advertising across YouTube, mobile gaming, and influencer marketing. It works with brands and agencies to place video campaigns alongside content that fits the brand and its standards, using its own AI models and contextual data.
The people who use the platform day to day are its ad operations teams, who build campaigns and monitor how they perform. Two of our engineers work on it: one on the web front end, and one who has been on the backend for years and by now largely keeps it running alone.
The people who use the platform day to day are its ad operations teams, who build campaigns and monitor how they perform. Two of our engineers work on it: one on the web front end, and one who has been on the backend for years and by now largely keeps it running alone.
The Challenge
The tool the adops teams used had been built up over years, and it showed. The front end ran on an old bundler with a stack of outdated libraries, the code structure had drifted, and the whole thing was written in JavaScript with no type safety.
On top of that, Precisify wanted to add serious new capability, AI modules for finding and vetting channels, without destabilizing the parts of the tool people relied on every day. On the backend, core functionality was being moved to Java, alongside new features for mobile-app advertising, influencer campaigns, and video.
On top of that, Precisify wanted to add serious new capability, AI modules for finding and vetting channels, without destabilizing the parts of the tool people relied on every day. On the backend, core functionality was being moved to Java, alongside new features for mobile-app advertising, influencer campaigns, and video.
Our Approach
We started with an audit of the existing front end, mapped where it was slow or fragile, and designed an architecture that new modules could plug into without breaking the logic already there. The JavaScript codebase is being migrated to TypeScript as the work goes, and the outdated libraries have been updated. That gave us a base stable enough to build the AI features on.
What We Built
The substance of the work is two AI modules, both built on Precisify's own models.
Content Recommender
A chat-driven way to find channels. A specialist types a brief in plain language, something like a campaign for a given product in the UK aimed at teenagers aged 10 to 18, and the model returns a list of matching YouTube channels. From there, the conversation continues. Narrow the list, or add a new angle, through follow-up messages. Once the list looks right, the specialist selects the channels they want and exports them straight into an ad campaign, where the ads will run.
Content Auditor
A checking tool for channels that are already in play. It takes a list of channel IDs, or a ready-made ad group with channels already in it, and runs them through filters for brand-safe topics, gender, and age. The output shows, channel by channel and video by video, what passes the chosen filters and what doesn't, so a campaign can be cleaned up before it goes live.
Outcome
The platform is in active use by Precisify's ad operations teams. Development continues, with the focus now on further AI modules and finishing the move to TypeScript.
Tech Stack
Frontend: TypeScript, React, MUI, Mantine, TanStack Query, React Hook Form, Zustand, Recharts, Vite, xlsx | Backend: Java 17 (migrated from Java 11), Spring Boot 3.5.3 with Spring Cloud microservices, Spring Web/WebFlux, Data JPA, Security, Actuator, Messaging | Databases: Amazon Aurora, MongoDB, Amazon Redshift (analytics), Hibernate, Flyway | Cloud & ML: AWS SageMaker, Amazon Comprehend, S3, DynamoDB, Secrets Manager | External APIs: YouTube Data API v3, Google Sheets API, Google OAuth, Tubular | Security: Spring Security OAuth2, JWT, SAML2 SSO, Bcrypt | Tooling: Springdoc OpenAPI / Swagger, Apache POI & Commons CSV, Lombok, MapStruct, Aho-Corasick
Feel like your project would win from our work? Let's talk!
sales@smartificial-ai.com
© 2026 Smartificial
AI Solutions Cases
