Build the AI feature—and the workflow around it.

Contributions to AI-generated story quizzes and profile-driven themes in an existing Telegram Bot and Mini App platform.

My role: AI feature & full-stack contributor
Status: Team project · Contribution overview
Built with: FastAPI · Celery · React · TypeScript

The team context

Serendip-TG is a Telegram Bot and Mini App platform centred on human connection. I contributed specific AI features within an existing team codebase. The platform’s wider matching and social experience is team work; this overview focuses on the generation workflows, editing tools, and interface integration I worked on. That distinction matters because the value of a contribution is in the behaviour it adds and the problems it resolves.

Story generation and review

My work included an AI-generated story quiz flow and the surrounding generation pipeline. Generated content needs a path from a request to a draft, review, and publication. I contributed admin draft editing, generation-status listing, and status polling so an operator can understand where a request stands and work with its output. The result is a more complete feature than a single call that returns generated text.

Keeping background work reliable

Generation can take longer than an ordinary request and may complete while other work is happening. My contributions included releasing database transactions during external generation, handling concurrent completion, and improving the background workflow’s status handling. I also worked on CI test fixes around the feature. These changes focus on reliable coordination between the API, worker tasks, persistence, and the interface that needs to report progress to a person.

Profile-driven interfaces

I also contributed to UI theme generation through an asynchronous pipeline and corrected theme-cache invalidation. This work connects structured profile information to a personalised interface while keeping generation separate from ordinary page interactions. The portfolio presents these as contributions to the existing product, with no private source or user profiles exposed. Working in a shared codebase reinforced the importance of explicit feature boundaries, reviewable changes, and fitting new AI behaviour into established product flows.

Discuss this project