Jump to

Actions

Projects

Notes

↑↓ move · ↵ open · esc close
Alex López Gómez

Leli

/ˈle.li/ noun what friends call Alex López

Founding engineer at mafer AI.

Born in , based in , in love with startups since 18. I build AI for real-world operations, think in systems and ship fast. I write about AI, go-to-market and startups on my Substack.

LinkedIn ↗

Experience

  • mafer AIFounding Engineer

    2026 - Present

    Building the operating system for specialty chemicals R&D: lab and instrument data, historical formulas and regulatory intelligence in one place.

    mafer.ai ↗
  • Prosper AIAI Engineer

    Jan 2026 - Aug 2026

    Agentic AI platform for healthcare to power administrative tasks from patient scheduling to insurance verification and patient billing. Backed by a16z, Emergence Capital, Base10, and Y Combinator.

    getprosper.ai ↗

    STT · MCP · LLMs · Voice AI

    • Added a parallel STT pass on the live call pipeline to catch and correct language mismatches in real time. Rolled out from a feature flag to company-wide use across 70,000+ production calls, at 96% correct-switch accuracy and a 5% cut in hang-ups.
    • Shipped the first client rollout of a journey platform unifying order flow across separate automated calls, synced to the client’s system of record: 58% of orders automated without human intervention (vs. 45–50% baseline), saving 8 hours/week in week one.
    • Built the end-to-end workflow-analysis engine for an internal MCP-based assistant that lets AI PMs inspect and debug call-resolution logic conversationally without engineering support. Live in production and adopted by 60% of AI PMs.
  • Coinscrap FinanceAI Engineer

    2025 - 2026

    Fintech that helps banks and insurance companies get more out of their transaction data through machine learning: a production ML system trained on millions of data points, serving enterprise clients.

    coinscrapfinance.com ↗

    TypeScript · React · Next.js · MongoDB · n8n · AWS · Docker · Python

    • Built an AI-powered natural-language analytics dashboard on a multi-tenant MongoDB: ask a question in plain Spanish, get a chart. Schema-first retrieval cut token usage by ~89%, and a query cost-scoring layer auto-approves, monitors or blocks every query before it hits the database.
    • Automated the CTO’s daily email workflow with an AI pipeline that reads incoming mail, extracts structured information, posts summaries to Notion and sends digests to the right stakeholders. It stayed in use because its rules were explicit and debuggable.
    • Built an internal two-agent prompt-engineering tool that interviews non-technical teammates, turns their requirements into structured JSON and compiles production-ready prompts with auth and data standards built in.
    • Took part in the company’s ISO 27001 security review, going through every system one by one with the whole team.
  • FreelancerAI Automation Engineer

    2023 - 2025

    Developed AI-driven automation and chatbot solutions that automates lead generation, while streamlining SME operations to save 3.5 hours weekly (avarege) through end-to-end inbound and outbound strategies.

    LLMs · Make · LangChain · n8n · Voiceflow

  • UDIASpeaker/Instructor

    2024 - 2025

    I taught +10 classes on AI systems development and go-to-market strategies for the Udia community, one of the largest communities in Spain.

    udia.es ↗

    Public speaking · Teaching

Projects

4 selected

Quorum

01/04
About
An open-source RAG evaluation framework powered by a Council-of-LLMs architecture. Routes each test case through an adaptive strategy layer that runs parallel AI judges (OpenAI, Anthropic, Gemini) or falls back to lightweight deterministic checks based on risk scoring, then aggregates results into a single verdict with full cost and latency breakdown.
Stack
Node.js, React, MongoDB, LLM, RAG

Smart Home Assistant System

02/04
About
Multi-agent smart home assistant built with Jason that simulates a domestic environment with coordinated robot, owner, and auxiliary agents for tasks like medication delivery and home automation.
Stack
Jason, Java, Multi-Agent Systems

Taxi Prediction

03/04
About
End-to-end ML system that forecasts hourly NYC taxi demand using LightGBM, feature stores, and real-time Streamlit dashboards.
Stack
LightGBM, Hopsworks, Streamlit, Python

Credit Scoring

04/04
About
Machine learning model to predict customer credit risk based on financial history, using features like debt ratio, monthly income, and late payment history to estimate default probability.
Stack
Scikit-learn, Pandas, Numpy, Pytest, Python

Notes

Show 11 more
All notes on Substack ↗