jandresgarcia.dev
Rev. 2026.09
ENES
Portrait of Julian Garcia

Julian Garcia, AI Engineer and Fullstack Software Engineer

AI Engineer with a senior fullstack core. I build LLM products that hold up in production.

Download CV Email me
See what I build and run

Medellín, Colombia
English · Spanish
Now: ChenMed · EApply · Clara

Application notes

The AI products I designed, built and run end to end, the platform they run on and the open-source tools I ship. Real apps, real runs.

AN-01

EApply Tailors a CV to a job post with an LLM, without inventing a single fact.

Case study EApply Closed beta · eapply.jandresgarcia.dev ↗

Video 1. EApply, product overview · 0:30

60 → 91

ATS fit on a real run. Every figure in the CV untouched.

How it works

  1. CV + job post

Engineering notes

  • Shows an honest ceiling, the best score reachable without inventing experience, instead of a fake 100.
  • Multi-model by design: a fast model extracts, a quality model writes, a vision model reads scans and judge models grade the output.
  • Production controls: per-model rate limits, a daily AI budget, and every call traced with its cost.

LLM pipelineMulti-model routingLLM-as-judge evalsGuardrails

Next.js · Fastify · Postgres · BullMQ · Redis · Typst · OpenTelemetry

AN-02

Clara A personal finance agent where the model never does the math.

Case study Clara In daily use · private app

Video 3. Clara, product overview · 1:20

Zero guessed numbers

A deterministic validator blocks any number the model did not get from a tool.

0

Numbers guessed by the model. A deterministic validator blocks any figure that did not come from a tool.

How it works

  1. Question
  6. Answer

Engineering notes

  • The LLM explains and compares. A deterministic engine owns every balance, projection and score.
  • Three model tiers picked by a deterministic router, with per-role fallback and a daily spend cap.
  • A live eval with real models (60 cases) caught failures the hermetic suite missed; fixed with server-side guards.

AI agentTool callingMulti-model routingLive evals

Next.js PWA · Fastify · SQLite · Drizzle · Zod · 3 model tiers · Docker on ARM64

AN-03

JAG HQ The self-hosted platform where every product is built, deployed and run.

Case study JAG HQ Private · tailnet only

Video 5. JAG HQ, product overview · 0:42

1 VPS

Every product, its CI runners, Plane and the control center on one Always Free ARM server, behind Tailscale.

How it works

Engineering notes

  • Backstage rebuilt as a keyboard-first control center: a ⌘K command palette, j/k navigation and seven windows for the wall, projects, host, logs, docs and tools.
  • Live ops from the source: logs streamed from Docker through a read-only socket proxy, host checks every 5 minutes, uptime monitors and daily restic backups.
  • Built for agents: skills, SDDs, handoffs and house rules live in the repo. Every change is a Plane ticket, one worktree per agent, merged only on green CI. Architecture diagrams are generated from the code with Archify.

Platform engineeringBackstageSelf-hosted CI/CDObservabilityAgent workflow

Backstage · Docker Compose · Caddy · Tailscale · Cloudflare Tunnel · Plane · GitHub Actions (self-hosted) · Uptime Kuma · restic · OCI ARM64

AN-04

Agentville A pixel-art office where every Claude Code session works, and you approve, answer and message them all from one desk.

Case study Agentville Open source · macOS · agentville.jandresgarcia.dev ↗

Video 7. Agentville, product overview · 1:03

0

Network calls. Sessions, orders and approvals travel as files on your own machine.

How it works

Engineering notes

  • Every session is an office: the Lead and its subagents at their desks, idle agents in the lounge, and agents that need your OK waiting in line at your desk, where A and R approve or reject.
  • One panel for every session: a usage screen with spend, context and tool calls, each agent's plan, steps and conversation, and a message field that reaches any session or subagent.
  • An optional team flow: the Lead triages each change small, medium or large by risk, and nine plugin agents (Secretary, Product, spec-writer, builders, four reviewers and an approver) take it from spec to merge.

Open sourceClaude CodeMulti-agentDeveloper toolsLocal only

TypeScript · Claude Code plugin (hooks, agents, skills) · Claude desktop app · HMAC-SHA256 · MIT

Features

  • Production LLM systems: RAG, vector search and multi-model orchestration
  • Security, observability and cost engineered in, not bolted on
  • 4+ years shipping Java, Spring, Kafka, GraphQL, React and Next.js
  • Proven in US healthcare: cut load times 30%+ on doctors' tools at ChenMed
  • End to end, from system design to deploy

Typical applications

  • AI SaaS products and internal copilots
  • Adding an AI layer to an existing microservice platform

Key characteristics

ParameterValue
Professional experience4+ yrs
Load-time reduction, ChenMed>30%
Current roleChenMed, since 2023
AI products runningEApply, Clara
Own platformJAG HQ, self-hosted
Open sourceAgentville, Claude Code mod
AI layerwhat I build now
LLM productsEApply, Clara
RAG & vector searchRetrieval, embeddings, vector DBs
OrchestrationMulti-model routing, agents
Production AI: security · observability · cost
Fullstack core4+ years in production
Product UIReact, Next.js, TypeScript
Services & dataJava, Spring, Kafka, GraphQL, MongoDB
PlatformDocker, Kubernetes, CI/CD, Datadog
Figure 1. Functional block diagram. The AI layer runs on a senior fullstack core.

Availability

Open to senior AI Engineer and Fullstack roles, and to AI product builds for founders.

Download CV (PDF)EmailLinkedInGitHub

Revision history

PeriodRole
2023 – nowSoftware Engineer · ChenMedInternal apps doctors use to care for seniors across the US. An architecture refactor cut load times by more than 30%.
2022 – 2024Senior Frontend Developer · Void.ggLed key features of an enterprise microservice platform in Next.js, with code reviews across the frontend.
2021 – 2022Full Stack Developer · Kurius ChocolateE-commerce in React with Stripe payments and a MongoDB data layer; designed in Figma.
2020 – 2021Software Engineer · Mantum SASProduct interfaces in TypeScript and Next.js, built from Figma designs.