Multi-Player Games improved via Discord

TL;DR

Contracted for a few weeks to optimize Accord, a product that transforms how game studios handle community feedback.

The Product Journey

The three founders behind Accord came from gaming tech, and they’d already built something remarkable: Layer Licensing, the world’s largest IP licensing marketplace for gaming. After Layer got acquired, they turned their attention to another problem they’d seen firsthand in the gaming industry: community management at scale.

Community managers at game studios read through thousands or millions of messages daily trying to identify patterns, prioritize issues, and surface feedback to development teams.

Accord’s solution is elegant: ingest Discord messages, run them through an ML pipeline, and surface organized insights on a dashboard. The product turns chaos into clarity. Instead of reading everything, community managers get a high-level view of what their players actually care about.

My Role

They brought me on as a contractor with a clear brief: optimize the slow parts.

Subtopic Classification

The ML pipeline was already classifying messages by topic (e.g., “Bugs,” “Story Feedback,” “Feature Requests”). I built the subtopic layer on the UI. For example, “Bugs” might break down into subtopics like “UI Issues,” “Crash Reports,” “Lag/Performance,” all ML-generated. Users could edit, merge, or delete topics and subtopics to customize the taxonomy for their game. This gave studios more granular control, not just “people are reporting bugs,” but “47% of bug reports are about multiplayer lag.”

Frontend Optimization

The dashboard was handling millions of rows of data (Discord messages from massive gaming communities), and the frontend was slow. Really slow.

I optimized queries, improved data fetching patterns, and reduced render times. Working with Neon Postgres (serverless PostgreSQL), Azure Databricks and Next.js/TypeScript, I identified bottlenecks and refactored components to handle large datasets without choking.

GraphQL Data Layer

One of the key tasks was creating a unified data access layer using Hasura DDN (Data Delivery Network).

I helped build the integration between Hasura DDN and Neon Postgres via GraphQL, giving the app a single, consistent data source. This cleaned up the architecture and made it easier to query data across the ML pipeline, database, and frontend without manual data wrangling.