Davit Papikyan

Developed novel ML-driven product attributes to enhance merchandiser catalog quality and product discoverability. Refactored and optimized core ML codebases, and contributed to backend development by designing and implementing API endpoints.

Researched, developed, and deployed end-to-end ML pipelines for visual analytics, including model architecture development, training, optimization, and production integration. Reduced inference latency of an object detection model by 60% through TensorRT quantization, enabling real-time deployment on edge devices.

Researched adversarial attacks in self-supervised learning and introduced the backbone attack targeting image representations. Engineered software to analyze attack transferability across models.

Developed and deployed a production-grade AI voice assistant using Dialogflow API. Fine-tuned NLP models to automate transit-style support flows and integrated chat and voice channels.

Built scalable data pipelines and developed performance metrics to measure the success of WhatsApp Business. Created real-time dashboards to visualize and monitor key product insights.

MSc in Mathematics in Data Science from the Technical University of Munich, with a thesis on diffusion models as latent priors in generative models.

BSc in Informatics and Applied Mathematics from Yerevan State University, with a thesis on numerical calculation of a mathematical model of blood vessel deformation.