From Code to Direction: Deriv’s VP of Engineering on Rebuilding the Software Development Pipeline Around AI

Why It Matters

By centering AI in the dev‑ops workflow, Deriv cuts coordination overhead, accelerates delivery, and curbs technical debt—an approach that could reshape fintech engineering efficiency.

Key Takeaways

• Engineers act as directors, AI executes code, tests, and documentation.
• Unified steering documents let AI enforce consistent standards across teams.
• AI‑driven QA agent auto‑detects issues and generates permanent tests.
• Pipeline stays deterministic despite AI’s non‑deterministic nature.
• Roadmap aims for fully automated deployment from intent to production.

Pulse Analysis

Fintech firms are scrambling to sprinkle AI onto legacy development stacks, often creating fragile add‑ons that slow rather than speed delivery. Deriv’s strategy flips that script by designing its pipeline from the ground up with AI as the execution engine. Engineers define high‑level intent, standards, and quality gates, while AI consumes these directives to write, test, and document code. This separation of intent and execution eliminates traditional hand‑offs—code reviews, manual QA, and infrastructure provisioning—allowing teams to move from concept to production in a single, streamlined flow.

The technical backbone hinges on two principles: deterministic pipelines and embedded governance. Deriv packages steering documents, best‑practice templates, and quality checkpoints directly into repositories, ensuring AI‑generated output adheres to organization‑wide standards without relying on individual memory. Simultaneously, the system enforces deterministic CI/CD stages, counteracting AI’s inherent non‑determinism and guaranteeing repeatable builds. A concrete illustration is the AI‑driven QA agent that autonomously detected a Sri Lanka service error, logged a detailed ticket, and generated regression tests that permanently safeguard against the fault. Such self‑healing capabilities reduce on‑call fatigue and improve uptime across the platform’s global footprint.

Looking ahead, Deriv aims to close the final gap—automated deployment—so that a developer’s natural language description triggers end‑to‑end code generation, testing, and live release without manual infrastructure steps. If successful, this model promises a new engineering paradigm where skill, not syntax, becomes the primary programming language. For the broader fintech ecosystem, Deriv’s AI‑first pipeline offers a blueprint for scaling speed, consistency, and reliability while keeping technical debt in check, potentially setting a new industry standard for rapid, high‑quality software delivery.