Luxembourg

Alejandro Diez Crespo

I lead data science and AI for Amazon's fulfilment business in Europe. I build AI agents that people use every day, and I measure what they change.

Portrait of Alejandro Diez Crespo
Voice, English or Spanish, up to 3 minutes. It's an AI, not me.

Or start with:

Audio is processed by OpenAI. Alejandro gets a written transcript of each call.

From a delivery station in Madrid to AI in Luxembourg

Eleven years on one line. Select a stop to see what I did there.

  1. Telecom engineeringUniversidad Europea · University of Hertfordshire, 2015–2020

    Dual BEng and an MSc, with theses on speaker recognition and on machine learning for unstructured data.

    • Dual BEng Universidad Europea and University of Hertfordshire
    • MSc thesis on machine learning for unstructured data
    • C1 English and AWS Certified AI Practitioner
    All the details
    • MSc Telecommunications Engineering, Universidad Europea (2019–2020).
    • Dual BEng, Electronics & Digital Communications / Telecommunications Engineering, University of Hertfordshire · Universidad Europea (2015–2019).
    • AWS Certified AI Practitioner · Cambridge Advanced English (C1)
  2. Software internNTT Data, 2018

    Angular frontend components.

    • Angular frontend components
    All the details
    • Built Angular frontend components.
  3. Risk advisoryDeloitte, 2019–2020

    Risk-assessment tooling and a custom compliance application for enterprise clients.

    • Risk tooling for enterprise clients
    • Compliance app built to measure
    All the details
    • Worked on site with enterprise clients; delivered risk-assessment tooling and a custom compliance application.
  4. Area ManagerAmazon, 2020–2022

    Launched a delivery station and led 60+ people; workflow changes adopted across Europe.

    • 60+ people led at a new delivery station
    • 15% time saved, adopted across Europe
    All the details
    • Launched a delivery station, led 60+ people and redesigned workflows for 15% time savings adopted across Europe.
  5. Senior Data AnalystAmazon Logistics, 2022–2024

    Capacity-planning app still used by 40+ program managers, about 7 hours saved a week.

    • 40+ program managers still use the planning app
    • ~7 h saved a week, ~40% more productive
    • 300+ delivery stations forecast in nine countries
    All the details

    Built and supported production planning software and forecasting models for European operations.

    • Shipped a React and Python capacity-planning application still used by 40+ program managers: ~7 hours saved per week, ~40% higher planner productivity.
    • Built an AutoGluon forecasting ensemble across 300+ delivery stations in nine countries.
    • Ran planning pipelines over 600+ raw KPIs; added Meta Prophet forecasting for ~4% better plan-versus-actual accuracy.
  6. Head of Data Science & AIAmazon, 2025–now

    Built the team from zero; its agent stack serves a 150+ person organization.

    • 72% of seller defect cases solved on first contact
    • 90 → 5 min per coaching case, with ~2,100 sellers
    • AI Trailblazer 2026 selected from 500+ people
    All the details

    Head since Jan 2026 · Senior Data Scientist Feb–Dec 2025

    Hands-on leader building and operating applied AI systems for internal account-manager teams and the external sellers they support.

    • Shipped a Bedrock multi-agent defect-resolution pilot in about two months: ~72% first-contact resolution, ~2 minutes per case versus a 30-minute target, ~100 users.
    • Built a Bedrock and Redshift seller-coaching platform used for ~2,100 sellers: coaching time down from ~90 to 5 minutes, return rate improved ~11%, ~94% of investigations run autonomously.
    • Built multi-tenant RAG and evaluation infrastructure with Bedrock Knowledge Bases, OpenSearch and MCP, using golden datasets and hallucination-aware LLM-as-judge rubrics.
    • Won Amazon FBA's 2026 AI Trailblazer award (shared with one co-builder, selected from 500+ people).
    • Turned one-off builds into a reusable agent stack for a 150+ person organization; prompt caching cut token cost by ~90%.
    • Built the team from zero: ran 30+ technical interviews, then built alongside four engineers.
  7. TopTechSide project, 2025–now

    Runs the whole product: frontend, backend, billing, analytics and deployment.

    • End to end frontend, backend, billing, analytics and deploys
    • Realtime voice mock interviews
    • Hardened verified webhooks, bounded retries, rate limits
    All the details
    • Built and operate an AI career platform end to end alongside a full-time role: CV analysis, mock interviews, salary intelligence and cover letters with Claude, OpenAI Whisper and text-to-speech.
    • Own the React/TypeScript frontend, Python and Convex backend, Stripe billing, PostHog analytics, Vercel deployment and every external API integration.
    • Hardened production with verified idempotent webhooks, structured model output, bounded retries with model fallback and per-IP rate limits.

Hiring for applied AI?

hola@alejandrodiez.com

Or ask the AI first: it knows my work and answers in under a minute.

Talk to an AI about my work