I'm an M.Sc. Computer Science graduate from Sapienza Università di Roma with a strong focus on AI/ML, computer vision, generative AI, and production-oriented software engineering.
I enjoy building AI systems that go beyond experimentation — from RAG applications and LLM-powered systems to computer vision models and cloud-deployed ML APIs.
My research has focused on Vision Transformers and continual learning, and I have published research in an IEEE conference.
AI & Machine Learning
Python · PyTorch · TensorFlow · Scikit-learn · Keras · Hugging Face · Computer Vision · Deep Learning · Vision Transformers · Continual Learning · Generative AI · LLMs · RAG · AI Agents
Software Engineering
FastAPI · REST APIs · React · Git · Data Structures & Algorithms · Object-Oriented Programming
Cloud & DevOps
Docker · AWS ECS · AWS Fargate · CloudWatch · Render · Linux
Data & Infrastructure
NumPy · Pandas · Pinecone · Vector Databases · Gemini API
An end-to-end Retrieval-Augmented Generation chatbot that answers questions over FastAPI's official documentation while providing source citations with every response.
- Modular four-stage pipeline: ingestion → embedding → retrieval → generation
- Recursive paragraph/sentence-aware chunking with Unicode normalization
- Gemini API embeddings with task-specific query/document embeddings
- Pinecone vector database for managed retrieval
- FastAPI backend with
/chatAPI - React frontend with per-answer source citations
- Self-indexing deployment that automatically builds the knowledge base on a fresh deployment
- Deployed backend designed around a 512 MB RAM constraint
Tech: Python · FastAPI · React · Gemini API · Pinecone · RAG · LLMs
An AI-powered document comparison platform that detects semantic changes between document versions rather than relying only on character-level diffs.
- Sentence-level change detection: added, removed, and modified
- Distinguishes numerical changes from wording changes
- Built semantic embedding and cosine-similarity pipelines
- ~840 ms warm-request latency on CPU-only inference
- Async FastAPI REST APIs with request validation
- Containerized using Docker
- Deployed on AWS ECS/Fargate
- Monitored with AWS CloudWatch
Tech: Python · FastAPI · Sentence Transformers · PyTorch · Docker · AWS ECS/Fargate · CloudWatch
My Master's research project at the Vision Lab, Sapienza Università di Roma, focused on reducing catastrophic forgetting in Vision Transformers during continual learning.
- Developed PyTorch training and evaluation pipelines
- Implemented adapter-based continual learning and Elastic Weight Consolidation (EWC)
- Evaluated sequential classification across 5 Split CIFAR-100 tasks
- Achieved 90.59% average accuracy
- Achieved only 2.01% average forgetting
- ~97% relative reduction in catastrophic forgetting compared with a CNN baseline
- Benchmarked against ResNet-18 and naive ViT fine-tuning
- Validated results through cross-seed evaluation
Tech: Python · PyTorch · Vision Transformers · Computer Vision · Continual Learning
A computer vision system for automated industrial metal surface defect detection using the NEU dataset.
- Custom CNN using TensorFlow/Keras
- SVM and Random Forest models using HOG features
- Image preprocessing and augmentation pipelines
- Hyperparameter optimization with GridSearchCV
- 97.22% test accuracy using the CNN
- Compared deep learning against classical computer vision approaches
Tech: Python · TensorFlow · Keras · OpenCV · Scikit-learn · Computer Vision
Evaluating Pretrained Vision Transformers Under Continual Learning
Research focused on adapter-based continual learning and Elastic Weight Consolidation for mitigating catastrophic forgetting in Vision Transformers.
Evaluated multiple machine learning approaches including Random Forest, SVM, Decision Trees, and Neural Networks for air pollution prediction.
M.Sc. Computer Science
Sapienza Università di Roma · 2026
Relevant coursework: Deep Learning & Applied AI · Machine Learning · NLP · Computer Vision · Distributed Systems
B.Tech — Computer Science & Engineering (AI & ML)
Sharda University · 2023
AI/ML Engineering · Generative AI · RAG · LLM Applications · Computer Vision · Machine Learning Research · AI Systems