Lisbon, Lisbon, Portugal
I’m a backend engineer who builds systems that scale and deliver measurable impact. At Tracklib, I: • Scaled the platform 3.5× to 70,000+ users by optimizing Django/Celery pipelines and AWS infrastructure • Built a recommendation engine (Elasticsearch + ML) that improved music discovery for thousands of users • Engineered a global payment system processing royalties across 40+ countries • Cut AWS S3 storage costs by 30–40% through compression and archival strategies Before that, I researched privacy-preserving synthetic data generation at Syndata (Stockholm), implementing Transformer-based models that outperformed traditional architectures on accuracy and privacy benchmarks. I’m strongest when I can own a problem end-to-end: understand the business need, design the data/ML layer, and ship reliable backend services that handle real traffic. Currently looking for Senior Backend Engineer or ML Engineer roles (remote/hybrid, EU‑friendly time zones) where I can own backend architecture for payment or data‑intensive systems and apply AI to real‑world products
• Scaled platform 3.5x to over 70k users through Django/Celery optimizations I led • Built ingestion pipelines handling 10k+ tracks daily, accelerating catalog growth • Developed recommendation engine (ElasticSearch + ML) that boosted music discovery • Engineered global payment system processing transactions across 40+ countries • Optimized AWS S3 storage, cutting costs by 30% through compression strategies
• Developing and testing MVPs for AI tools: chat assistants, workflow automation, and backend intelligence. • Exploring how AI can streamline real business operations through lightweight, maintainable prototypes. • Focused on bridging practical backend engineering with applied AI research. • Building open-source components and templates to accelerate AI product development.
• Researched synthetic data generation for privacy-preserving AI training. • Designed and compared deep learning architectures including RNNs, LSTMs, and bidirectional attention models. • Implemented and evaluated a Transformer-based model (similar to ChatGPT’s core architecture), achieving superior performance and privacy preservation compared to traditional methods. • Delivered a validated approach that enabled companies to train models without exposing sensitive datasets.