Admir Krnic

Five years of production data engineering, an MSc in Informatics from TU Munich, and first-author AI research alongside the day job. The specialism is making unreliable data trustworthy: contradictory, multi-source data with no canonical identifiers and no ground truth.

DATA ENGINEERING (current, pharmaceutical market analytics)
Owns the regulatory and supply chain data domains end to end, petabytes per sync across billions of rows. Rebuilt the drug data parsing module from scratch, now at 70 to 100 percent extraction success. Redesigned the marketing authorisation dataset from flat to nested and rebuilt the pipeline around it, taking mapping accuracy on priority markets close to 100 percent. Unified US supply chain data across three sources with no shared identifiers, resolving hundreds of thousands of entities. Expanded regulatory coverage by more than ten countries.
Stack: Python, SQL, PostgreSQL, Trino, dbt, SQLMesh, Temporal, LLM-based processing.

AI RESEARCH (concurrent, part-time, neural audio watermarking)
First author on the first systematic study of backward compatibility in zero-bit neural audio watermarking, a requirement under the EU AI Act Article 50 Code of Practice. Introduces a loss that restores true negative rates to 99 to 100 percent against out-of-chain embedders at no cost to accuracy. Under review at IEEE WIFS 2026. Also first author on contrastive learning work reaching 95 to 99 percent bitwise recovery across 25 attack types.

MSc Informatics, TU Munich, grade 1.9. BSc Computer Science, University of Montenegro, GPA 9.92 of 10. Strongest where requirements are ambiguous and nobody can hand over a correct answer.

Seeking fully remote mid to senior roles worldwide. Based in Montenegro, works on CET, available via employer of record or as a contractor.