Carlos Díaz
Software Engineer
Based in Copenhagen. I work on applied ML research end-to-end, from model design to the pipelines and compute that run it.
Get in TouchWork Experience
Denmark Technical University
Research Software Engineer
Department of Applied Mathematics and Computer Science - Dynamical Systems Section
November 2025 — Present- Build and maintain a data pipeline processing 30 years of weather, road survey, and maintenance data across millions of road segments in Denmark, extending to European cities including Lisbon, Zagreb, Zilina, and Tartu in collaboration with local road authorities.
- Implement road deterioration forecasting using Dynamic Bayesian Networks and Palmgren-Miner fatigue models, integrating short-term and long-term climate forecasts with road condition assessments to optimize maintenance scheduling.
- Engineer feature extraction from heterogeneous temporal and spatial datasets to feed probabilistic and deterioration models.
- Deploy and run experiments on a NVIDIA GPU cluster, transitioning local workflows to accelerated compute.
Python 
PyTorch 
Kubernetes 
Docker Resiplus ADDInformatica
Software Engineer R&D
October 2023 — January 2025- Developed a C#/.NET pipeline (4 microservices) that auto-generated ERP documentation from Git commit parsing code Deltas against DSL-defined XML schemas, extracting domain logic, and producing context-aware Markdown and PDF manuals.
- Developed HL7 FHIR integration services acting as a data exchange gateway between the internal ERP and external healthcare systems.
C# 
.NET 
HL7 FHIR Education
ETH Zürich
MSc in Computer Science (SEMP Scholarship)
Zürich, Switzerland
2026 — 2027University of Copenhagen
MSc in Computer Science
Copenhagen, Denmark
2025 — 2027Relevant coursework: Machine Learning, Deep Learning, Online and Reinforcement Learning, Advanced Computer Systems, Recommender Systems, Big Data Systems, Search Engines.
Polytechnic University of Valencia
BEng in Informatics Engineering
Valencia, Spain
2021 — 2025Academic excellence distinctions in Industrial Formal Methods, Databases and Information Systems, and Model-Driven Software Development.
Featured Projects
Search Engine with LLM-based Retrieval
2026 - University of CopenhagenBM25 and LM retrieval pipelines built with PyTerrier over 152K documents and 10K queries, analyzing preprocessing trade-offs between index size, latency, and ranking quality. Added LLM-based query expansion (hypothetical-answer prompting) and bi-encoder/cross-encoder re-ranking, quantifying each method's quality gain against its latency cost — cross-encoder re-ranking gave the best quality-per-millisecond trade-off.
Python 
Terrier 
Transformers Stateful FaaS Platform
2026 - University of CopenhagenA stateful Function-as-a-Service platform (Microsoft Orleans, Kafka, Redis) with multi-step workflow orchestration and exactly-once processing over Kafka's at-least-once delivery, using idempotency keys and atomic Redis transactions. Benchmarked under concurrent load, showing throughput scaling from 50 to 110 req/s while p95 latency dropped from 225ms to 44ms as Kafka partitions increased from 1 to 10, with contention analyzed via compare-and-set retry rates under skewed traffic.
.NET 
Kafka 
Redis 
Microsoft Orleans Hybrid Recommender System
2025 - University of CopenhagenCollaborative filtering (KNN, SVD via scikit-surprise) and content-based (TF-IDF, GloVe, DistilBERT via HF transformers) models on a 98% sparse Amazon dataset, with NumPy/Pandas feature pipelines. A Reciprocal Rank Fusion ensemble of SVD + TF-IDF lifted coverage from 20% to 40% while preserving ranking quality. Also explored Qwen 3.5-4B item descriptions as a metadata replacement.
Python PyTorch 
Pandas Transformers Atheres Fit
2025 - Past projectA cross-platform fitness platform connecting athletes and trainers, with a web dashboard where trainers generated AI-tailored workout plans that auto-adjusted loads per athlete. MVP shipped to the Apple App Store.
Flutter 
Next.js 
PostgreSQL 
About Me
Hi, I’m Carlos Díaz, a research software engineer based in Copenhagen, working across ML and software engineering. I’m currently at the Technical University of Denmark, where I build probabilistic models and data pipelines that forecast road deterioration across European cities. I’m also doing an MSc in Computer Science at the University of Copenhagen, and in 2026 I’ll be at ETH Zürich on a SEMP scholarship. Before research, I shipped production services in C#/.NET, including HL7 FHIR healthcare integrations. Outside of work, I enjoy prototyping product ideas with friends.
