Work

Airbnb Superhost-recognised monthly rentals business in Bangkok managing 100+ properties. As Founding Engineer, I own the technology platform end to end, from architecture and infrastructure to production operations.

• Designed and built the core ERP platform on GCP, supporting day-to-day business operations across the organisation.
• Re-architected external integrations around an event-driven model, replacing synchronous processing with dedicated Go workers to improve fault isolation and eliminate request timeouts under load.
• Identified duplicate processing caused by at-least-once delivery guarantees and implemented idempotent pipelines with retries and exponential backoff, eliminating manual recovery.
• Introduced systematic load testing with k6, uncovering and resolving throughput bottlenecks. On modest infrastructure, the platform sustains 50+ req/s at P95 < 1.5s.
• Built CI/CD, monitoring, logging, and alerting from scratch using Prometheus, Grafana, Loki, and Sentry, improving incident detection and response.
• Secured the platform with least-privilege IAM, managed secrets, and application-level rate limiting.
• Built an AI-assisted messaging platform combining intent classification, tool calling, and RAG pipelines with human-review checkpoints for low-confidence responses.

Stack: TypeScript, Go, PostgreSQL, Redis, OpenAI, GCP

Built an OCR system for Mon, a UNESCO-classified vulnerable language. As no public datasets, OCR models, or supporting toolchain previously existed, the project began with collecting and validating training data before building the models and deployment infrastructure from the ground up.

• Developed two OCR models: a lightweight on-device model (~11.4M parameters) and a higher-capacity server model (~54M parameters).
• Supports deployment across Web (WASM), Android (NNAPI), iOS (CoreML), and backend environments through ONNX export.
• Built a human-in-the-loop feedback system that continuously collects validated samples and improves recognition accuracy over time.

Stack: Python, PyTorch, ONNX, CoreML, SvelteKit, Go

Geospatial platform used by organisations across Africa and Asia for community mapping and land-use planning.

• Built the ingestion layer that normalised data contributions from multiple countries, each with different formats, update schedules, and access requirements.
• Maintained and operated the platform throughout its full lifecycle, from implementation through long-term production support.

Stack: React, PostGIS, Mapbox GL, MongoDB, Redis, GraphQL, GitLab CI/CD, Linode

Led the redesign of a legacy biodiversity platform, simplifying the architecture while improving reliability and reducing operating costs.

• Migrated the system to a Next.js + Baserow architecture, reducing infrastructure costs by 33%.
• Rebuilt data pipelines to reduce operational failures and manual intervention.
• Introduced backup and recovery capabilities that were absent from the previous system, improving resilience and disaster recovery readiness.

Stack: Next.js, Baserow, PostgreSQL, GeoServer, Vercel

Built in partnership with Kontinentalist and the Global Initiative Against Transnational Organized Crime.

• Designed and built a global monitoring platform that aggregates wildlife trafficking intelligence from multiple sources under strict security and compliance requirements.
• Engineered the ingestion layer to normalise incident data across sources with differing formats, structures, and update frequencies.

Stack: SvelteKit, PostgreSQL, Vercel