maya.log

$ whoami
maya reyes — full-stack engineer, mern specialist, lisbon

Full-stack systems that hold up in production.

Five years turning product specs into MongoDB schemas, Express APIs, React interfaces and Node services that stay online after the demo ends.

What I actually own at each layer

Four letters, four different jobs. This is the ledger, not the buzzword grid.

layerowned here
MongoDB schema design & indexing — document shape, compound indexes, aggregation pipelines that don't scan the collection.
Express REST & auth middleware — route contracts, request validation, session and token handling.
React state & component architecture — data flow that doesn't re-render the whole tree on every keystroke.
Node queues & deployment — background jobs, CI, the process that keeps running at 3am.

Selected work

Three shipped systems, formatted the way I actually track them — as a log, not a gallery.

a7f3c1e ledger-sync MongoDBExpressReactNodeWebSockets Cut month-end reconciliation from 3 days to 40 minutes.

problem

Finance teams were reconciling bank feeds against internal ledgers by hand, in spreadsheets, once a month — every mismatch found a week too late.

result

Now running at 12 mid-market companies. Reconciliation that took three days of an accountant's time now finishes itself in 40 minutes, unattended.

architecture

client (react)api (express)queue (bullmq)db (mongodb)

screen

2b9d40a routewise MongoDBRedisExpressReact Reduced route-planning time by 68% across a 40-vehicle fleet.

problem

Dispatchers were building next-day routes manually against a paper map and a gut feeling, re-planning by phone every time a driver ran late.

result

Route plans that took 45 minutes to build now take 14. Late-day re-routing runs live, from the same optimizer, on the dispatcher's own screen.

architecture

client (react)api (express)cache (redis)db (mongodb, geo-indexed)

screen

9e17f88 signalboard NodeWebSocketsMongoDBReact 2.4M events/day at sub-200ms dashboard refresh, p95.

problem

Support leads needed queue health and agent load on screen in real time, not in a report that ran at 6am against yesterday's numbers.

result

Now the system this whole site's architecture section is about — see below.

architecture

client (react)gateway (node/express)pub/sub (redis)db (mongodb, sharded)

screen

// zoom: signalboard

One dashboard, four hops, no dropped frame.

client
react
api gateway
node / express
pub/sub
redis
db
mongodb, sharded

signalboard's clients hold a websocket to a Node/Express gateway that fans every event out through Redis pub/sub before it lands in a sharded MongoDB cluster — the path that lets 2.4M events a day resolve into a chart update in under 200ms at p95, on a support floor that is watching it live.

currently taking one new contract for Q4 2026 lisbon, UTC+1 replies inside a business day, usually a few hours

Send the spec

Tell me what you're building. I'll reply with questions, or a plan.

POST /contact HTTP/1.1

Content-Type: application/json

{

}

no spam, no newsletter — just a reply.

Full-stack, end to end — and built to hold.