Led development and maintenance of ZAURU, an enterprise ERP system serving 2,000+ daily active users with 500+ requests per minute.
Managed and mentored a development team for custom web applications integrated with the ERP & CRM APIs.
Set up and operate a Hasura GraphQL service over the company ERP, used by associated web applications to query and create data against the core system.
Designed and delivered 6 full e-commerce sites built with Webstudio, integrated with Payload CMS (fully customized and extended) and CRM APIs, using Payload CMS GraphQL to query and create content and data for those web applications.
Built and maintained cloud infrastructure on AWS: containerized services on ECS, asynchronous job and email queues with SQS, relational databases on RDS, and open-source services hosted on Lightsail, alongside Lambda and S3.
Own day-to-day CI/CD for web applications using GitHub Actions, from build and test through deployment to production.
Architected and built offline-first web applications using Electric SQL and TanStack DB, enabling full functionality without an internet connection.
Adopted and integrated modern tooling including React Router (framework mode), shadcn/ui, TanStack Query, and TanStack DB across new and existing projects.
Conducted extensive client-facing requirements gathering sessions with diverse stakeholder profiles, demonstrating strong emotional intelligence and professional composure.
Designed the team's AI-agent delivery workflow: coding agents work in isolated sandboxes on dedicated GitHub branches, exercise existing features through layered tests in their own environments, and open pull requests that pass an AI code-review agent before a final human review. Shared test and production systems stay untouched until a polished PR is ready to merge.
AI Highlight: Integrated Google's gemma_q4f16_embedding model to generate embedding vectors for all ERP items; built a search API that vectorizes user queries at runtime and computes similarity scores against the item vectors enabling semantic, fuzzy-text search across online stores beyond exact keyword matching.
AI Highlight: Designed an isolated, human-in-the-loop delivery pipeline for coding agents using Cursor, GitHub Copilot, OpenCode, and a self-hosted Ollama server running Qwen: each agent works in its own sandbox and GitHub branch, runs layered tests against existing features in a disposable environment that never touches shared staging or production, then opens a pull request that is first reviewed by an AI code-review agent and finally by a human. Agent PRs arrive already tested and reviewed.