Gael Valdivieso · AI & automation consultant

I build automations that real employees use every day — and I measure what they change.

Final-year Engineering and Management student at the University of Exeter and founder of Gael AI Consulting. Replaced a factory's paper maintenance process with an AI-assisted app used every day, then sold the client the next phase. Before that, automated reporting at FedEx, official representative in Ecuador (Grupo Entregas), saving ~4 hours of manual work a day.

Let's talk → See the work
>80%
faster to log a maintenance request with AI
~4 h/day
of manual reporting eliminated
5,199 / 0
records processed, errors — first test with real data
1 role
vacant role made unnecessary by an AI-assisted app
Case studies

What I built, and what it changed

01 – 04 · result first
01
AI-assisted app · in production
>80%
faster to log a maintenance request
1 role
vacant coordinator role made unnecessary
304
machines in the live catalogue

AI maintenance platform for a printing and packaging plant

Gael AI Consulting · Jul 2026 – Present
Result

Replaced a vacant maintenance-coordinator role. Logging a maintenance request is >80% faster. Used every day by the plant manager and the technician.

Problem

Corrective maintenance at a ~550-employee printing and packaging plant (304 machines) ran on paper.

What I built

An installable web app (PWA) covering the whole workflow: request → diagnosis → quotation → approval → resolution. Role-based login, device revocation, role-targeted push notifications and self-updating deployments.

How it works
  • Voice dictation with automatic text clean-up.
  • Photo analysis that suggests each request's priority and classification.
  • Voice-dictated closure reports turned into structured data, with automatic time tracking and a formatted closure email.
  • Connected to the company's legacy FileMaker ERP: live 304-machine catalogue, two-way sync of requests, and an offline queue with automatic retry for unreliable plant connectivity.
Business side

Scoped, priced and sold a phased digitisation roadmap. Phase 1 approved by the client in Sep 2026.

In progress

Automated supplier-quote intake: inbound emails are verified, AI extracts amount, currency and terms, quotes are matched to the right machine and unmatched ones flagged for review.

Built and tested · deployment in progress
Next.jsReactTypeScriptTailwind CSSSupabase (Auth, Postgres, RLS)FileMaker Data APIOpenAI (speech-to-text, vision, extraction)ResendWeb PushVercel
02
Automated reporting · used daily
~4 h/day
of manual reporting eliminated
5,199 / 0
real records processed / errors
39
branches with daily visibility

Operations control tower for a national delivery network

FedEx, official representative in Ecuador (Grupo Entregas) · Jun – Jul 2026
Result

Eliminated ~4 hours of manual reporting a day. Management now has daily visibility of at-risk and delayed shipments across a 39-branch network. Used every day since launch.

Problem

An office employee consolidated the daily shipment report by hand.

What I built

First, 9 VBA tools that cut the report to a few clicks. Then a fully automatic control tower: the report arrives by email and is ingested automatically, validated before any write, de-duplicated and logged. It includes a KPI dashboard, an alert centre, risk-sorted shipment tracking with full per-shipment history, and 30-day trends by branch.

How it works
  • 5,199 real shipment records processed with zero errors in the first test with real data.
  • Branches detected and created automatically.
  • 11 users in 2 roles; each branch sees only its own data.
  • Plus a Power BI monitoring dashboard (Power Query, DAX, Row-Level Security).
Excel / VBANext.jsTypeScriptSupabase / Postgres (Row-Level Security)Cloudflare Workers (email ingestion)RechartsVercelPower BI
03
Guided web app · in use
4
bank clients covered
17
delivery rules encoded

Delivery guide app for bank-card couriers

FedEx, official representative in Ecuador (Grupo Entregas) · Jun – Jul 2026
Result

Couriers across Ecuador no longer need to call the office to ask how to deliver a bank card.

Problem

There was no written guide for bank-card deliveries; couriers had to remember the rules of each bank.

What I built

The first guide: a step-by-step web app covering 4 bank clients. The courier answers a few questions (delivery type, product, recipient) and gets the exact documents and warnings required, encoded as 17 delivery rules.

04
Personal project · retired
341
openings tracked
250
companies

Job-search automation pipeline

Personal project · Sep 2026
Result

Tracked 341 openings across 250 companies automatically, with a daily email digest instead of checking job boards by hand.

What I built

A scheduled Python pipeline that classified my application emails (read-only Gmail), pulled openings from job APIs and company job boards (Greenhouse, Lever, Ashby), de-duplicated them and refreshed every 3 hours, with a dashboard on Vercel.

Retired · since replaced
PythonGmail APIREST APIsGitHub ActionsVercel
Skills

What I work with

Applied AI
Speech-to-textVision classificationDocument data extractionLLM tool calling (OpenAI, Gemini)AI-assisted development (Claude Code)
Automation & development
PythonTypeScript / Next.js / ReactSupabase / PostgresREST APIs & webhooksCloudflare WorkersGitHub ActionsVercelLegacy ERP integration (FileMaker Data API)
Data
Advanced Excel (VBA)Power BI (Power Query, DAX, Row-Level Security)SQL
Business
Scoping and pricing projectsBusiness casesRequirements gatheringClient management
Background

Operations first, then automation

Education
BEng (Hons) Engineering and Management · University of Exeter, UK · 2024 – 2027
Predicted
First Class Honours
Languages
Spanish (native) · English (native/bilingual) · French (B1)

Process before automation. In 2025 I cut a packaging plant's folding-carton production cycle by 24 hours by finding a machine bottleneck and leading senior management to invest in a second machine.

Relevant coursework: Industry 4.0 · Modelling of Engineering Systems (Python) · Manufacturing Systems · Operations Management.

Contact

Let's talk.

valdiviesogael@gmail.com ↗ linkedin.com/in/gael-valdivieso ↗
Let's talk →

If the button above doesn't open your mail app, use "copy email" and paste it into whichever client you use.