Miguel Oliveira

AI & Machine Learning Engineer. Portuguese, based in Copenhagen. Building ML systems at Spradley AI and Contimo. MSc in AI at DTU.

I write about machine learning, AI engineering, and the messy parts of taking models from a notebook into something that actually runs for real users.

migueloliveir8@gmail.com · GitHub · LinkedIn · CV (pdf) · RSS

Writing

• 2026-05-30 Notes on the future: AI, SaaS, and the markets that will bleed
  AI won't kill SaaS. It will compress per-seat pricing, bleed single-dev DevTools, and bleed early-stage technical VC. The rest of the market just changes shape.
• 2026-04-04 Agent harnesses don't transfer
  Everyone now agrees that the harness contributes more than the model. What people are not saying out loud is that the harness pattern we converged on was shaped for coding, and copying it into other domains is making most non-coding agent products worse than the chatbots they were going to replace.
• 2026-02-15 Claude Council: when one agent isn't enough
  A Saturday hack that turns Claude Code skills into a multi-agent debate. The chat interface is the wrong shape for hard decisions.
• 2025-10-08 Engagement surveys are theatre
  We used to do qualitative research. Then scale broke it and we built surveys instead. LLMs let us go back, if we get the pipeline right.

All posts →

Path so far

School and work, most recent first.

• 2026–present
  
  CTO

  Spradley AI, Employee Research Platform

  AI-moderated employee research platform. Replaces the classic engagement survey with structured AI interviews that go deep on the why behind a number, and turns thousands of conversations into themes and signals leadership can act on. EU-hosted, GDPR and EU AI Act aligned. Technical lead across the interview agent, the analysis pipeline, and the product.

  • LLM orchestration for structured interviews; vector retrieval over org-specific context to keep each conversation grounded in the company's own language.
  • Async analysis workers turning raw transcripts into themes, signals, and reports.
• 2026–present
  
  AI/ML Engineer

  Contimo, RTB Advertising Platform

  ML for an RTB / native advertising platform serving publishers, brands, and agencies. Work spans editorial-side content selection and campaign-ops monitoring.

  • Pulse: trending-content discovery for publishers. AI pipeline ranks top articles and generates localised headlines across languages.
  • Campaign Alerts: real-time campaign health monitoring. Delivery forecasting plus anomaly detection on pacing.
• 2025–present
  
  MSc Eng., Artificial Intelligence and Algorithms

  Technical University of Denmark (DTU)

  MSc in Copenhagen. Coursework in deep learning, optimisation, probabilistic models, and algorithm design.

• 2025–2026
  
  AI/ML Software Engineer

  Revenya, Investment Banking

  ML systems for corporate lending and real-estate analysis. Production focus: reproducibility, auditability, and correctness of model outputs feeding lending decisions.

  • GroundTruth: multi-agent real-estate market intelligence. Autonomous research, analysis, and PDF report generation.
  • Credit-scoring pipeline automating risk assessment for ~$2M projected monthly loan volume. Approval times: days → minutes.
• 2024–2025
  
  AI/ML Engineer

  Napps, E-commerce Platform

  Owned the AI side of a no-code app builder for e-commerce. Shipped recommendation, automation, and lead-scoring systems on an AWS MLOps stack.

  • Hybrid Graph Neural Network recommendation engine for real-time personalisation.
  • AWS MLOps pipeline with automated CI/CD for model training and deployment.
  • LangChain agentic workflow parsing unstructured e-commerce data for client onboarding.
  • NLP-driven lead-scoring system prioritising high-intent leads.
• 2023–2024
  
  Erasmus+ Exchange

  University of Hradec Králové (FIM)

  Semester abroad at the Faculty of Informatics and Management. Software engineering and AI coursework in English.

• 2021–2025
  
  BSc in Computer Science

  Porto Polytechnic Institute of Engineering (ISEP)

  Four-year programme in computer science. Core CS plus ML, software architecture, and systems work.

  • Capstone: dual-model recommendation architecture (Temporal Graph Networks + evolutionary-optimised KNN). Later published at ICITS'26.
  • Coverage: data structures, OS, networks, databases, software architecture, ML.

Publications

• 2026 A Dual-Model Architecture for E-commerce Recommendation Systems
  ICITS'26 (Springer Proceedings). Context-aware dual-model: Temporal Graph Networks for discovery, evolutionary-optimised KNN for comparison.

Full CV →