L1
Frontend
Interfaces people actually use — including offline, in the field.
- Next.js
- React
- PWA · offline + QR scanning
- iframe SDK design
- Tailwind CSS
Senior Full-Stack Engineer/Los Angeles, CA
Senior full-stack engineer with 7+ years architecting and shipping internal platforms end-to-end. Go & Python on the back end, Next.js on the front, Azure underneath — from greenfield database design through production support.
End-to-end. A request passes through every layer below — and I’ve designed, built, and deployed all of them.
Interfaces people actually use — including offline, in the field.
The contract everything else depends on. Lately: Go, stdlib only.
The right store for each job — transactional, audit, cache, graph.
Containerized, CI/CD-delivered, SSO-secured, globally routed.
These platforms are internal and proprietary — so here’s the engineering, not the screenshots: the constraint, the architecture, and the call that mattered.
Internal LLM Platform
A multi-tenant LLM assistant platform embedded across 5+ internal apps.
Many internal apps wanted an embedded assistant — but each has its own knowledge base and its own access boundaries. One shared bot could not be allowed to leak one app’s data into another.
Isolation by signed identity, not by deployment. One platform serves every app, but a per-client JWT makes data bleed structurally impossible instead of a thing you hope nobody misconfigures.
Self-Initiated · Auth Infrastructure
A centralized JWT identity service for the whole app portfolio — proposed and built solo.
The same LDAP / auth boilerplate was duplicated across every application, with shared secrets spread around the portfolio. Nobody owned the problem — so I proposed the fix and took it from prototype to production.
Asymmetric over shared-secret. Publishing a public JWK set means apps can verify tokens without ever holding a secret to distribute, leak, or rotate in lockstep. Security that gets easier to operate, not harder.
Offline-First Field PWA
Bi-annual asset verification for 5,000+ staff across 15+ country missions — usable with no network.
Field staff in low-connectivity locations have to verify physical assets twice a year. A connection cannot be assumed — the app has to work fully offline and reconcile later.
Design for the worst network, not the average one. Treating offline as the default case — not an error state — is what makes the tool actually usable for the people it’s built for.
More, owned end-to-end
Production Latency Debugging
Vehicle, driver & trip management — one API serving a web admin and a partner mobile app.
Believe the users, then measure. The slowness was real but invisible from HQ — the fix lived in edge routing and caching, not application code. Knowing where to look saved chasing the wrong layer.
Integration · Reporting at Scale
Multi-country daily balance entry with automated Citi Bank API sync and executive dashboards.
One surface, two audiences. Entry and executive reporting share a single source of truth, so the numbers leadership sees are the numbers finance just entered — no export step to drift out of sync.
Same Stack · Owned End-to-End
10+ internal platforms on the same Django / Go + Next.js + Azure stack, owned from design to deploy.
A consistent stack is a feature. Standardizing on Django / Go + Next.js + Azure means every new platform inherits the same auth, deploy pipeline, and operational playbook — so shipping the tenth is far cheaper than the first.
How I work matters as much as what I ship — especially in the age of AI-assisted development.
AI-assisted code goes through a dual-model generation and review cycle — one model writes, another critiques — before a human ever signs off. Speed from the tools, judgment from the engineer.
Comprehensive coverage across unit, integration, and smoke tiers — with testcontainers and evaluation harnesses for LLM apps — so regressions surface before users do.
Architecture decisions, trade-off calls, and final verification stay human-owned. The tools accelerate the work; they don’t make the decisions.
Greenfield design through Azure deployment and production support — I’d rather spend the time getting the architecture right than rushing to ship something I’ll be paging myself about later.
Architect and lead developer on 10+ internal web platforms used by 5,000+ employees across 15+ country missions — spanning finance, operations, IT, and HR. Drove the team’s adoption of Go for new services and own end-to-end delivery: database design, API, front end, containerization, and deployment via Azure DevOps to Azure Container Apps.
05 — Contact
Open to senior full-stack & platform roles · open to relocation · visa sponsorship needed