Botan AHMAD
Data Scientist
Based in Paris, France
About
I am an experienced data scientist with a strong background in AI and machine learning, having collaborated with companies such as Klark, Delfox, Deepn, Airbus, Thales, and Nexter. My expertise includes designing advanced deep learning models for computer vision and natural language processing, developing reinforcement learning algorithms, and creating simulation environments using Unreal Engine. I am proficient in Python, Java, C++, and Rust, and have experience deploying solutions on AWS and GCP. I am passionate about integrating AI solutions, creating customized models, and optimizing technical workflows to deliver tangible results.
Skills
Latest projects
Deep Reinforcement-Learning-Based Air-Combat-Maneuver
I designed a hierarchical decision-making framework for autonomous air combat maneuvers using deep reinforcement learning. The framework consists of two layers: a strategic combat policy for selecting optimal maneuvers and a control policy for executing them through precise aircraft inputs. This approach achieved an 85.7% success rate in within-visual-range combat scenarios and reduced training time by 13.6% compared to traditional reinforcement learning methods, showcasing its efficiency and operational effectiveness.
Experience
2024
ML Engineer at Klark
Developed advanced RAG systems and LLM-based agent workflows to automate customer support tasks, including answering inquiries and drafting emails. These solutions were deployed across over 50 client CRMs, leveraging techniques such as Chain-of-Thought (COT), self-reflective RAG, LLM re-rankers, and ReACT frameworks. Fine-tuned large language models (LLMs) and adapted retrievers for specialized tasks and client-specific applications, utilizing both open-source and proprietary models. Additionally, contributed to a range of NLP tasks, including FAQ generation, message clustering, ticket categorization, and text translation, showcasing expertise in diverse linguistic challenges.
2022-2024
ML Engineer & RL Researcher at Delfox
Conducted advanced research in Reinforcement Learning, focusing on state-of-the-art algorithms and techniques to drive innovation. Developed and simulated embedded systems and fighter aircraft models to enhance flight dynamics and decision-making capabilities. Researched and implemented efficient Transformer architectures, such as Efficient Transformer and Fast Attention, to improve model speed and scalability. Reviewed and implemented cutting-edge research papers in Reinforcement Learning, staying at the forefront of AI advancements. Established robust data pipelines to process and manage large-scale datasets, ensuring efficient data flow and accessibility. Designed realistic simulation environments in C++ using Unreal Engine, enabling comprehensive testing and training scenarios.
2020-2022
Co-Founder & CEO at Deepn
Led the development of an algorithmic trading platform that enabled traders to create custom trading strategies using a proprietary programming language, Flow. Designed and implemented Flow, optimized for high-speed execution and user-friendly application in trading scenarios. Oversaw end-to-end product development, from conceptualization to production, including backend architecture and user interface design. Conducted research on Deep Learning models to improve predictive accuracy and enhance trading algorithms. Prioritized speed optimization to ensure real-time trading capabilities with minimal latency.
2020 - 2022
Research Engineer Intern at Airbus Defense & Space
Contributed to research and technology (R&T) projects, focusing on innovative solutions. Worked on embedded systems, including IoT devices and autonomous drones, to develop cutting-edge applications. Designed and implemented Deep Learning and Reinforcement Learning models, specializing in video anomaly detection and computer vision tasks.
2018 - 2020
ML Engineer at Sniffy
Developed a model to extract emotions and behaviors from video streams, leveraging advanced computer vision techniques. Conducted state-of-the-art research to analyze sentiments and emotions from facial expressions. Utilized an unsupervised model to interpret and refine the results of the captured emotional data, ensuring deeper insights and accuracy.
Education
2020-2022
Master Degree – Data Science at ESGI
Specialized in advanced data science methodologies, focusing on deep learning, statistical learning, and reinforcement learning. Gained expertise in Bayesian optimization, time series analysis, and other cutting-edge techniques. Developed the skills to design, implement, and optimize data-driven models for diverse applications in AI and machine learning.
2017 - 2020
Bachelor Degree – Computer Science at ESGI
Completed a comprehensive program covering foundational and advanced topics in computer science, including algorithms, data structures, discrete mathematics, linear algebra, probability, calculus, and numerical methods. Gained expertise in operating systems, computer architecture, and low-level programming with C/C++. Developed a solid understanding of the theory of computation, equipping me with the analytical and technical skills to solve complex computational problems.