Project Overview
Collide, a professional network for the energy sector, underwent a full platform rebuild focused on modernizing its legacy Ruby on Rails codebase. The project introduced AI-powered chat alongside an enterprise-grade solution, real-time messaging, and a gated application system, delivering a scalable, engaging platform that helped secure a $5M seed funding round and positioned the company for accelerated growth.
Technologies Used
- Ruby on Rails
- Hotwire
- Hotwire Native
- Stimulus
- Tailwind
- JavaScript
- Microsoft Azure
- PostgreSQL
- Customer.io
- Stream
- Redis
Service Type
Full Stack Development, Mobile App Development, AI Development, Product Engineering
The Challenge
Collide was gaining traction as a niche professional community platform, especially in the oil and gas sector, but under the hood, it was brittle. The application had been stitched together with legacy code, broken media, and inconsistent user experiences across devices. The platform lacked modern communication tools, had no intelligent onboarding controls, and its visual design felt dated. Collide needed more than bug fixes. It required a full product transformation: a better interface, stronger community infrastructure, and new AI-powered workflows to match its ambitions.
At the same time, leadership envisioned evolving the platform into a curated, high-quality network rather than a wide-open forum. That meant introducing a gated application system to vet members, enrich profiles from external sources, and maintain quality at scale. Meanwhile, user engagement features like chat, personalized feeds, and an AI assistant for industry-specific queries had to be built from scratch. The challenge was not just to modernize the experience, but to turn a fragile, inherited codebase into a resilient and intelligent platform that professionals could rely on every day.
The Solution
The project began by stabilizing a fragile platform that had been operating on a deteriorating Ruby on Rails foundation. Collide’s backend logic was fragmented, the frontend design lacked cohesion, and many key user experiences were broken or outdated. From content repairs to AI integrations, every layer of the system had to be rethought with long-term growth, usability, and extensibility in mind.
The first phase of work focused on technical triage and UX recovery. The inherited codebase was plagued with broken media references, such as video and image links that failed to render or load. Content sizing was inconsistent across screen sizes, and the layout behavior varied wildly depending on the device or browser. I resolved content-level issues, standardizing layout behavior and implementing a new media handling strategy across the platform. This included logic to dynamically fallback from broken links, lazy load heavier assets, and apply consistent rendering rules on all screen breakpoints. It ensured users could reliably view and interact with posts without jarring visual issues or long load times.
At the same time, I overhauled the visual language of the platform. Collide had grown in popularity, but its styling was dated and lacked a coherent system. I implemented a fully custom TailwindCSS-based styling layer, introducing design tokens, layout primitives, and reusable components that respected accessibility and modern usability standards. This wasn’t a simple redesign. It was a ground-up recreation of the user interface across all views, with a focus on clarity, information hierarchy, and mobile-first responsiveness. With these improvements, the platform experienced an uptick in retention and MoM user growth, attributed to smoother onboarding flows and a clearer path to community engagement.
Collide’s leadership wanted to evolve the platform into a more curated and high-trust environment, which required control over who could access it. To meet this need, I built a gated application system that allowed prospective users to apply for access, rather than sign up freely. Each application was enriched automatically with external data from the web, including LinkedIn profiles and other public-facing attributes scraped on the fly. Admins were given a lightweight review interface with auto-scored inputs, history tracking, and profile context to make fast, informed approval decisions. This system was not only effective at ensuring high user quality, but also built to scale with minimal moderation overhead.
Once the gatekeeping mechanism was in place, I turned attention to user engagement features. Real-time communication was a clear missing piece. I implemented a full-featured direct messaging system using the Stream Chat API, bringing modern messaging capabilities to the platform without the overhead of building a messaging engine in-house. This included support for persistent threads, read indicators, media attachments, and structured conversation histories. The introduction of DMs created a new behavioral layer on the platform.
Beyond the messaging layer, I rebuilt the feed system to reflect familiar community paradigms. Collide’s previous content discovery felt scattered and hard to follow. I reworked the structure of user-generated posts into a vertically scrollable feed that mimicked the rhythm of Reddit, including upvote/downvote dynamics, active threads, and lightweight content summaries. The goal was to create a self-sustaining engagement loop where high-quality posts surfaced organically and prompted discussion. Each feed interaction was tightly integrated with Stimulus controllers for instant updates and Hotwire streams for seamless updates across sessions.
A unique requirement of the platform was the introduction of a domain-specific AI assistant tailored to the oil and gas sector. This was not a generic chatbot. I implemented an embedded AI system that allowed professionals to ask natural language questions related to their work — such as technical standards, field operations, document drafting, and more to receive contextual answers. The assistant was trained and prompted specifically for the energy industry and became a core differentiator in how Collide served its user base.
Under the hood, the application was streamlined for maintainability and developer velocity. Everything was built using Ruby on Rails with Hotwire, Stimulus, and TailwindCSS, allowing for a minimal-JS but high-performance full-stack experience. I maintained a clear separation of concerns between views, components, and controllers. This allowed for rapid iteration and testing without compromising stability. I also introduced deployment hygiene practices such as feature flagging, seed data versioning, and structured migrations to reduce risk during releases.
Throughout the project, I worked with minimal oversight while remaining aligned with product leadership. The end result is a revitalized professional platform that enables curated membership, real-time messaging, AI support, and a community-first content model all built with maintainable code and designed for future expansion.
Collide now has the infrastructure and UX foundation to support hundreds of thousands of users and mobile-native versions without the overhead of a full rewrite. With high-quality users, rich profiles, intelligent features, and strong engagement loops, the platform is now a modern home for the energy workforce online.