I turn ambiguous operating problems into dependable products—from production APIs and data pipelines to AI-assisted workflows and the interfaces around them.
~30%
lower backend latency
Production optimization at Adaviv
900
simulation runs
Across 45 DRT scenario variants
10,078
failures analyzed
KORAIL predictive-maintenance research
AWS
Solutions Architect
Associate certification
Merged
security PR #1480
Upstream OpenHands contribution
Flagship work / proof of ownership
Live with real listing inventory and an agent partnership, with provenance and verification limits built into the buyer experience.
A live Belize property-discovery and operations platform built around trustworthy inventory, buyer qualification, and real market constraints.
A public, inspectable reference implementation for validation, authorization, event processing, idempotency, and observability.
A production-shaped FHIR R4 ingestion platform demonstrating secure asynchronous backend engineering with synthetic data.
One configurable codebase now serves two property brands with bilingual, structured guest guidance.
A bilingual booking and guest-concierge platform for Belize stays, with a public discovery surface and private post-booking portal.
900 simulation runs across 45 scenario variants, with learning policies improving wait-time and service-rate direction over the greedy baseline.
A reproducible simulation platform for evaluating demand-responsive transit dispatch and learning policies under varied operating conditions.
Experience / production + research
Production ownership at Adaviv and reproducible mobility research at KNUT anchor the work. The common thread is software that must behave under constraints, not just look convincing in a demo.
Full experienceAdaviv · Remote
Built and operated backend systems for an agricultural technology platform used in real production environments.
Korea National University of Transportation · South Korea
Designed reproducible research systems for mobility simulation and railway failure analysis.
Role fit
Production APIs, data models, integrations, performance, reliability, and the infrastructure around them.
LLM and ML features designed as accountable product systems, with evaluation and fallback behavior.
Working close to users to turn ambiguous operating constraints into deployed software.
End-to-end range when a small team needs technical judgment beyond a narrow layer.
Belize · UTC−6 · remote
Fragmented data, mobile-first users, informal workflows, and international teams shaped an engineering style that values clarity, resilience, and low operational overhead.
Selected engineering writing
Real estate in Belize doesn't start with a form submission. It starts with a Facebook post and a WhatsApp message — and the software has to meet buyers where they already are.
For Cayo Basecamp, the AI's constraint is the feature. A hospitality concierge that answers only from the host's approved guidebook — nothing from the open web — and uses failed answers to improve itself.
Demand-responsive transit is too expensive to experiment with in production. Here's how I built a simulation platform for comparing dispatch algorithms — and why reproducibility was a first-class requirement.
Open to remote roles
I am available for remote Backend and Applied AI roles, with range into forward-deployed and product engineering.