Commerce AIarXiv:2411.08392
18 min readNov 2024
Real-Time Purchase Intent Classification in Live Commerce Streams
Aryan MehtaPriya NairLucas OkaforSeo-Yun Park
Loading RunAsh
Loading RunAsh
RunAsh Research advances the science of AI-native live commerce — from intent models to agentic systems to real-time multimodal inference.
Intent classification, CTA optimisation, and conversion modelling for real-time selling environments.
Simultaneous processing of video frames, audio signals, and live chat to build unified scene representations.
Multi-step autonomous agents with tool use, parallel execution, and commerce-domain planning.
Low-latency translation and localisation systems tuned for live commerce vocabulary across 40+ languages.
Viewer cohort detection, churn prediction, and dynamic segmentation during live stream events.
Research into frictionless in-stream checkout, payment intent prediction, and fraud signals at scale.
Enter a commerce scenario and get a structured research-grade analysis — intent modelling, CTA strategy, conversion forecast, and risk factors — streamed in real time.
We are hiring across ML research, inference engineering, and applied science. Remote-friendly, research-first culture.
View open rolesNew papers, benchmark updates, and dataset releases — delivered when they ship.