Andrei Cojocaru

Ongoing research on improving LLMs’ mathematical reasoning and evaluation capabilities while reducing data requirements. Collaborating with Professor Sewon Min, pending submission to ICLR 2026.

Implementing an AI-driven solution to optimize AWS EC2 usage to achieve more than a 20% reduction in operational costs Leveraging AWS Cost Explorer API and Amazon SageMaker to automate the optimization, aiming for a 95% coverage rate Handling 1TB of data using SQL and ApacheSpark to enhance efficiency by 30%, and adapt to real-time operational needs

Optimized the functionality of generative AI in 5 projects, refining underlying models to enhance accuracy and performance Focused on the use of code refactoring, test and structured data creation to ensure efficient software development processes

Improved my knowledge of PenTesing, Threat and Vulnerability Monitoring and Management; Qualified for the DefCamp CTF final with my team, as well as completed HackTheBox challenges and TryHackMe learning modules; Assesed multiple companies' cybersecurity threats, proposing and implementing action plans to address them.

Developed new RPA processes, especially economics and financing tools for payments and invoices; Generated RPA processes for ERP systems and UiPath Tools, JavaScript, C++, and C#; Learned how to develop scrapping tools and automate the scrapping processes.