Konami Liu

I’m a Data Analysis Manager who blends analytical rigor with applied machine learning, from risk-focused financial work to production-minded ML/NLP engineering. I enjoy turning ambiguous business problems into measurable systems that improve outcomes.

In earlier roles across auditing and investor relations, I supported annual and IPO audits, built fundraising materials and roadshows, and communicated investment theses across funding rounds. Most recently, I identified inventory challenges and designed an intelligent production planning system, delivering 10%+ improvement in inventory turnover.

Alongside my industry experience, I completed an MSc (Distinction) in Speech and Language Processing and built end-to-end RAG QA and speech restoration research. My Tesla Model Y RAG QA system covered document parsing, hybrid retrieval, knowledge distillation, domain fine-tuning, quantization, and streaming API serving—achieving reported gains like +18.5% ROUGE-L and strong retrieval improvements.

I also researched lightweight alternatives to diffusion for speech bandwidth extension (8kHz → 16kHz), using iterative refinement with a time-conditioned architecture and evaluating with LSD and ViSQOL. I’m motivated by engineering that’s both technically sound and evaluation-driven, with clear metrics and reliable deployment paths.