Platform engineering - Zencoder: Inside AI, build once... scale everywhere
This is a guest post for the Computer Weekly Developer Network written by Andrew Filev in his capacity as CEO of Zencoder.
Zencoder is an AI coding agent platform designed to accelerate software development by automating tasks, enhancing code quality and improving developer productivity. It offers features like code generation, unit test creation, docstring generation, code repair and chat-based assistance, all powered by AI agents that understand a codebase.
Filey writes in full as follows…
The best platforms have elegant architectures, enabling an infinite number of use cases.
At my first company, Wrike, we started with a project management application. When we transformed it into a platform with the capability to customise the data model and workflows, something magical happened – millions of users found thousands of ways to utilise it that we never imagined. That transformation took us from a simple app to a category-defining company worth billions.
Now at Zencoder, I’m watching the same pattern unfold. We began with a coding agent – powerful on its own, but coding is only one step in SDLC. Instead of building separate coding agents, testing agents and review agents in a “50 first dates” style – where each tool is built in isolation, reinventing 80% of the work – we created a platform that enables dozens of interconnected use cases and can be customised for hundreds of different environments.
This isn’t just about technical elegance. It’s about compound value creation. Every new capability we add to the platform benefits all existing use cases. Every integration a customer builds can enrich their existing use cases. The platform becomes smarter, more capable and more valuable with each iteration.
The platform paradox
Here’s what seems counterintuitive in our age of AI-accelerated everything: taking time to build a proper platform actually makes you faster. Much faster.
Will Fleury, our head of engineering at Zencoder, puts it perfectly: “Treat the platform as a product, with your own engineers as its customers. This means having a clear value proposition, a published roadmap and tight feedback loops – exactly what you’d expect from any customer-facing SaaS product.”
The alternative isn’t simplicity, it’s chaos.
As Will notes, letting every team “roll their own infrastructure” creates far more complexity than a well-designed platform ever would.
The key is embracing industry standards rather than inventing bespoke solutions. They’re battle-tested foundations that let you focus on what truly differentiates your business. If you’re solving a common problem that has nothing to do with your core market differentiation, contributing to open source might serve you better than building proprietary solutions.
But there’s a critical balance to strike. Platform engineering isn’t about overengineering – it’s about enablement. The moment your platform makes implementing new use cases harder rather than easier, you’ve crossed the line from platform engineering into architecture astronautics. I’ve seen scenarios where, instead of simple REST endpoints and a handful of microservices, the solutions were overengineered to include GraphQL schemas across a dozen services.
The faster you move, the higher the chance that you will have to rethink some of your choices. For example, as the capabilities and failure modes of AI models have rapidly evolved over the last 12 months, we have had to rethink many internal details of our agentic platform and adopt new standards, such as MCP and A2A AI protocols.
AI-ready architecture
Platform engineering in 2025 requires a fundamental shift in thinking. Your platform must be consumable not just by developers, but by AI agents as well.
This isn’t futuristic speculation – it’s happening now. When you combine machine-readable contracts with coding agents, you get what Will calls “warp-speed SDLC.” Clean APIs, thoughtful documentation and well-structured interfaces are no longer just developer conveniences. They’re the foundation that allows AI to accelerate every phase of software development.
LLMs also provide an additional reason to embrace open source. When you contribute to open source, you’re not just building your employment brand or giving back to the community. Since all modern LLMs train on open-source code and documentation, you’re literally training the next update of frontier AI models on your platform. How cool is that? No DGX B200s required from you.
At Zencoder, we’ve built our platform with this dual consumption model in mind. Every API we expose, every documentation page we write, every integration we build – we ask ourselves: “Can both a human developer and an AI agent understand and use this effectively?” This “help me help you” philosophy has transformed how we think about platform design.
The human factor
Iurii Golikov, who heads engineering at Wrike, says “By enabling faster iteration and more responsive architectural adjustments, AI is helping companies keep pace with shifting market realities. The key is to balance the speed and flexibility offered by AI with the enduring need for robust, forward-looking platform architecture. So while AI is revolutionising how we build software, the foundational principles of good architecture remain as important as ever. Success in this new era will depend on our ability to harness AI’s power without losing sight of the strategic vision and technical rigor that underpin truly great platforms.”
This highlights a crucial paradox: as AI handles more of the routine, human elements become more important than ever. Alex Akimov from Monite learned this lesson the hard way: “Even though platform engineering is highly technical, the social interaction component is even more important and crucial for the success of platform teams. Great platform engineers must have high levels of empathy towards other engineers in the company and should genuinely think on how to make their lives better and their work more efficient.”
The platform-AI flywheel
We’re entering an era where platforms and AI create a powerful feedback loop, or flywheel effect.
Platforms make AI more effective by providing structured, well-documented interfaces for agents to work with. AI makes platforms faster to build and evolve by accelerating development and enabling new capabilities.
The modern platform engineering toolkit is evolving rapidly. Shared agents with a growing library of MCP servers are becoming as fundamental as traditional CI/CD pipelines. Platform teams now need to provide building blocks not just for human developers but for AI-augmented development workflows. For example, consider the following real-life scenario: one of the platforms where we deliver our agents is JetBrains IDEs and by default, LLM models often hallucinate when generating JetBrains code. We solved that dilemma two-fold: (1) by migrating some common blocks out of individual surfaces (eg Jetbrains, VSCode, CLI) into a common core platform (2) by an interesting AI recursion: we have used our team’s experience to date and our existing code to seed an AI agent that helped us build what we call our internal “Jetbrains Mega-Prompt” that enablea our AI coding agent to use the right APIs and architectural patterns, accelerating our Jetbrains development.
Consider the compound effect: A well-designed platform enables AI agents to be more effective. More effective AI agents help build better platforms faster. Better platforms enable more sophisticated AI use cases. It’s a virtuous cycle that rewards early adoption and thoughtful implementation.
The path forward
The choice facing engineering leaders today isn’t whether to invest in platform engineering – it’s how quickly they can build AI-ready platforms before the chaos tax becomes overwhelming. Every month you delay means more duplicated effort, more technical debt and more missed opportunities to leverage AI effectively.
But remember: platforms are ultimately about multiplication, not addition. In the AI era, this multiplier effect becomes even more powerful. When your platform can be consumed by both human developers and AI agents, every improvement cascades through your entire engineering organisation.
The companies that will thrive in this new landscape are those that bring together user empathy, strong architectural thinking and AI-first development processes. Start now. Build for both humans and machines. Embrace standards over custom solutions. And above all, remember that the best platforms emerge from solving real problems for real users.
