What I'm working on right now.

Inspired by nownownow.com. Updated when life shifts, not on a schedule.

NOW PLAYING · SPOTIFY

Nothing playing right now.

AT WORK · BLUE MEDICAL GUATEMALA

▶ ACTIVELY BUILDING

▸CCCV CRM — building the internal CRM for Blue Medical's commercial and medical operations. Laravel backend wired to the existing service mesh, centralizing doctor affiliations, commercial workflows, and scheduling data.

● OWNING IN PRODUCTION

Systems I built that are now on stable cadence — maintenance, hardening, and incremental improvements rather than ground-up new builds.

▸sofIA — primary author of a multi-agent voice system for healthcare scheduling, live in production. Built on ElevenLabs ConvAI + n8n with a FastAPI + Nuxt management platform. Recently tightened the validation layer around the LLM; now extending it to a WhatsApp confirmation channel alongside voice.
▸Payment Service — multi-gateway abstraction stable in production, recently extended with a new payment provider. Strategy Pattern keeps each integration scoped to its own DTO, strategy class, and job queue.
▸Invoice Service — Guatemala FEL e-invoicing through an authorized provider, with a Strategy Pattern for future provider swaps. SAP integration, multi-establishment support. Now expanding the tax-model seam to support Costa Rica.
▸BlueMeds Platform — medication subscription delivery with 10+ third-party integrations (ERP, support, messaging, healthcare middleware, AI calling).

◌ ON THE ROADMAP

▸More complex AI chatbots beyond sofIA's scheduling scope. Scope and architecture still being defined — likely multi-agent orchestration with stronger deterministic guardrails.

BUILDING ON THE SIDE · ACTIVE ALL PROJECTS →

LUMIRA: Real-time statusline for Claude Code & Qwen Code · TypeScript, zero runtime deps · v1.14 shipped with subagent-aware rendering and git-worktree fallback · published on npm · ~4k downloads/month
NOVA-ID: Self-hosted identity & SSO platform (OIDC via Ory Hydra) with a role/permissions demo API · actively hardening auth flows and audit logging
NIGHTWIRE: Compressed dark design system for developer tools · semantic tokens, intensity scale, Tailwind plugin · published on npm · v2 shipped
KOVIA: Smart pet adoption platform · connects rescue organizations with adopters via an automated scoring engine · NestJS + Nuxt 4 + Prisma + BullMQ
VITTBOT: Multi-agent crypto trading bot · 3 Claude agents + deterministic Risk Manager (10 hard-coded rules, never delegated to AI) · NestJS
CLARIFY: AI legal contract auditor (micro-SaaS, alpha) · Nuxt + Supabase RLS + OpenAI + Stripe credits + BullMQ
THIS SITE: The AI chat on this site is mine end-to-end · public chatbot grounded on my CV, now with multi-turn conversation memory and honest fit-boundary framing for comp/role questions · NestJS + Groq + Redis answer-cache · hardened against prompt-injection with a deterministic output sanitizer

RUNNING / SELF-HOSTED

This site, its API, and my side projects all run on infrastructure I operate myself — no managed PaaS.

▸A Hetzner VPS running everything in Docker Compose, behind Traefik as the reverse proxy and TLS terminator.
▸Self-hosted observability — metrics, uptime monitoring (alerting to Discord), and container log aggregation — so I see problems before visitors do. Admin dashboards sit behind a WireGuard VPN, not exposed to the internet.
▸GitFlow CI/CD: a release branch triggers an automated GitHub Release, image build, and zero-touch deploy over SSH. Merge to main, walk away.

LEARNING / EXPLORING

▸Multi-agent LLM orchestration — production-safe pipelines, deterministic safeguards over model choice, failure modes that don't require a rollback.
▸LLM evals and prompt engineering — output scoring, rubric-based benchmarks, systematic regression testing for AI pipelines. Making "does this actually work" answerable.
▸Payment network internals — network tokenization, card scheme authorization flows, PCI scope reduction. Goes deeper than most backend engineers need, but I'm already here.
▸Application security in practice — I ran an owner-authorized pentest against my own stack, then shipped the fixes: enforced CSP, real rate-limiting, prompt-injection defenses, and key-only SSH. Security as something you do, not just read about.

WHAT I'M LOOKING FOR

ROLE: Senior backend or tech lead at a company building real systems — payments, AI in production where the validation layer is the hard part, distributed services, the kind of work where the messy parts are the interesting parts.
LOCATION: Remote · UTC-6, works well with US hours.
STATUS: Not actively applying, but open to the right conversation. If what you're building sounds like it belongs on this page, reach out.