projects

Part of Team 6 building a security-focused monitoring stack for Chimera, UMass Boston's GPU cluster: Grafana dashboards driven by PromQL over metrics scraped by Prometheus (via the DCGM exporter) on a dedicated monitoring VM, deployed over SSH/SCP. I then researched the security trade-offs of exposing those metrics — documenting how an unprotected metrics endpoint opens the door to side-channel and Rowhammer attacks, and evaluating countermeasures like masking and self-correcting code.

Exploited C programs with buffer-overflow vulnerabilities to understand memory-safety issues — injecting malicious bytecode payloads with Python scripts to demonstrate the exploits, and using GDB to analyze memory layout, payload behavior, and potential mitigations.

A proof-of-concept keylogger in C, built to study system vulnerabilities and defenses. Implemented system hooks to capture and log keystrokes for research, exploring operating-system internals, file I/O, and the ethical implications of malware in cybersecurity.

Built and optimized a watchdog timer that resets a system when bugs are detected, hand-optimizing the C with x86 assembly for efficiency — deepening hands-on work with embedded systems, low-level programming, and VMware.

A machine-learning system in Python that analyzes text patterns in online posts to identify potential mental-health indicators. Applied NLP techniques — TF-IDF vectorization, Naive Bayes classification, n-grams, and FastText embeddings — to process and classify large datasets, building experience in pattern recognition and data analysis that carries directly over to monitoring and anomaly detection in security environments.

A Huffman encoding implementation in C exploring character encoding and data-compression fundamentals — building the encoding logic, an array-based encoding structure, and working through debugging and testing close to the metal.