Kacper Wikieł
ML systems that survive production. Compliance, industrial, enterprise.I build ML systems that work in production—not demos that impress in meetings and fail in deployment. Previously PwC Digital & AI. Led PLN 9.3M R&D project. Lecturer at University of Warsaw.
Specializing in adverse media screening for AML compliance, computer vision for industrial inspection, and enterprise RAG systems. I take on projects with clear metrics and full ownership—from architecture through deployment and maintenance.
What makes me different: I reverse-engineer proprietary formats others treat as black boxes, I bring Big4 methodology without Big4 overhead, and I don't do AI theater. If your ML project is stuck, stalled, or failing—I can help.
Selected Results
Adverse Media Engine — End-to-end AML screening system. NER + classification pipeline processing news sources for entity-risk detection. Production deployment for fintech client.
Ultrasonic NDT Pipeline — Reverse-engineered proprietary binary format from industrial inspection hardware. Built signal processing and CNN pipeline for defect classification. PERN S.A.
Enterprise RAG — Built core RAG infrastructure for P&G's internal GenAI platform. Custom chunking strategy, parallelization framework for batch inference.
Blockchain Compliance Platform — Tech lead on PLN 9.3M NCBiR-funded R&D project. Substrate-based chain with tamper-evident audit logs, KYC/AML rule engine.
Writing
Essays and notes. More coming.
-
Chunking strategies, retrieval quality, and the gap between demo and deployment.
-
NER, classification, and entity resolution for AML compliance systems.
-
When LangChain isn't enough: orchestrating multi-step LLM pipelines.
-
Spectrograms, feature extraction, and CNNs for industrial inspection.
-
When metrics matter and when they mislead.
-
From engineer to consultant: client management, scoping, and delivery.
-
Build vs buy, use case prioritization, and avoiding AI theater.
-
FastAPI, Docker, model serving, and monitoring.
-
AML, KYC, and the intersection of regulation and machine learning.
-
Binary analysis, pattern recognition, and building parsers from scratch.
Contact
For project inquiries: k.wikiel@gmail.com
LinkedIn ·
GitHub ·
CV
I work best with companies that have clear success metrics, direct decision-maker access, and real problems where ML is the right solution.