• Focused on enhancing the sample efficiency of Multi-Agent Deep Reinforcement Learning (MADRL) models.• Proposed a novel approach by integrating a Coordination Graph architecture into an off policy Learning algorithm SAC(Soft Actor-Critic).• Results: Achieved a sample efficiency improvement by 55% less interactions outperforming baseline methods, and by 3%in terms of average reward and 20% reduction in Convergence Rate.• Demonstrated these improvements through empirical validation in a traffic junction environment, with results suggestingpotential applications in real-world robotics.

• Built a voicebot to replace the Orange France call center ”Agent 118 712,” capable of receiving calls and processingcustomer requests.• Developed the Natural Language Understanding (NLU) pipeline using RASA (open-source conversational AI).• Integrated Asterix for VoIP functionality and Google’s API for Speech Recognition for converting speech-to-text andtext-to-speech.