• Built a full-stack AI app (React/Next.js, Python, Rust, Docker) using LLMs to analyze phishing emails, cutting SOC analyst review time by 67%

• Designed automated tests for data pipelines, ensuring reliability and scalability of AI solutions.

• Acted as the AI liaison across cross-functional data and cyber teams

• Conducted red teaming & adversarial attacks on generative models (LLaMA), identifying vulnerabilities in LLM outputs

• Engineered adversarial prompts and detection benchmarks for internal model safety tools

• Built and deployed a 100B parameter Bulgarian-language LLM using AWS SageMaker

• Enabled chatbot deployment & analysis of massive national-language corpora

• Empowered cross-border NLP applications for government and media use

• Delivered an anomaly detection pipeline via Splunk MLTK & Python, cutting detection time by 90%

• Built an XGBoost document classifier and deployed to production with full automation

• Created unsupervised document clustering using DBSCAN & TF-IDF, saving 40+ hours of manual labeling per week