Iman

Accelerated AI product development at Divar — Iran's leading classifieds marketplace (65M MAU, equivalent of leboncoin). Led data-driven annotation strategy combining active learning and automated tooling — 50% reduction in annotation time and faster ML deployment. Improved Farsi TTS/ASR F1 score, precision and recall by 35%. Diversified image processing capabilities with a precision image classification service for second-hand listings.

Led AI product at Cafe Bazaar, Iran's largest Android app store (70M MAU, B2B2C platform serving developers, brands, and advertisers). Designed and shipped Poshtibank Hooshmand, a conversational AI customer support agent with multi-turn dialogue, context retention, intent classification, and human-in-the-loop escalation — rolled out 24/7 to 15% of users, increasing positive feedback by 10%. Owned full ML lifecycle for Farsi TTS/ASR and NLP text analytics, including evaluation frameworks (WER, Precision/Recall/F1) and a co-developed RLHF pipeline — delivering +35% F1 improvement and 40% faster time-to-market. Managed a 40-person distributed team of ML researchers (including IOI medalists), engineers, data scientists, and designers.