WORK EXPERIENCE
-CYENS Centre of Excellence, Research Intern, Cyprus, September 2024 - January 2025
-Trained a policy using PPO and student teacher distillation to enable a robot to perform extreme movements such as jumping, crawling, and object avoidance using only depth images as input
-Designed and implemented terrains with moving obstacles in Isaac Gym, achieving a 2x simulation speedup by optimizing obstacle motion updates
-Huawei, Research Intern Part Time, Italy, March 2024 - June 2024
-Designed and implemented a pipeline to run simulations automatically, achieving a 6x speedup in dataset generation by creating a Python wrapper for classical and advanced solvers using Torch and JAX
-Replaced traditional linear solvers with CNN models including EfficientNet and ViT built from scratch in PyTorch, increasing speed by a factor of 1 million and improving accuracy by 200 times
-Created a UNet based diffusion model for metasurface design that generates results with a mean squared error of 0.01, outperforming the previous GAN based system by 40 percent
- Webranking, R and D Engineer, Italy, September 2022 - June 2023
-Optimized an existing SQL pipeline, reducing computational cost by 70 percent while maintaining predictive performance for client purchase probabilities
-Developed an internal chatbot that answers questions using internal documentation, vector databases such as Qdrant, LangChain, large language models, and embeddings from OpenAI and Hugging Face
-Built tools using Docker, React, and Django to support the SEO team with site indexing, API data recovery, AWS storage, and anomaly detection
-OPEN SOURCE
PettingZoo
Optimized and refactored tests for the standard multi agent reinforcement learning repository, reducing run times from 30 to 15 minutes
Identified and resolved bugs in various environments and improved documentation for environment configurations
-PROGRAMS AND COMPETITIONS
Participated in the ICPC SWERC competitive programming competition featuring top university teams across Europe in 2021 to 2022, only the 9 best students from each university were selected to compete
Attended the Mediterranean Machine Learning Summer School organized by researchers from Google DeepMind, accepted as a Masters student in a highly selective program with an 8 percent acceptance rate for Masters level applicants
