Academic and research work

Research question — How much predictive utility must be sacrificed to achieve meaningful, empirically-verified privacy guarantees when training credit-default models?

Contribution — Implemented DP-SGD training pipelines, ran membership-inference attacks across privacy budgets, and analyzed the resulting utility-privacy-fairness trade-offs.

Main result — DP-SGD stayed close to the non-private baseline across privacy budgets (AUC 0.857–0.861 vs. 0.867), while Local DP collapsed on the higher-dimensional dataset at tighter budgets (AUC 0.358 at ε=3) due to per-feature noise splitting. Membership-inference attacks stayed at ≈0.50 AUC across every configuration, including the non-private baseline.

Research question — How do sequential, shared-memory, distributed-memory and GPU parallelization strategies compare in runtime and energy efficiency for 1D Earth Mover's Distance?

Contribution — Implemented all five algorithm variants (sequential, OpenMP, MPI, CUDA, hybrid) and designed the profiling methodology used to compare them.

Main result — Measured on the University of Basel's SciCORE cluster: MPI+CUDA reached a 4,245× speedup over sequential for matrix multiplication (vs. 545× for CUDA alone), and CUDA reached 73× for 1D EMD's compute phase. Energy profiling showed OpenMP using ~91× less energy than MPI for merge sort at the same problem size.

Research question — Can a decentralized, gossip-based traffic and hazard dissemination system deliver operationally comparable utility to a centralized broker, while measurably reducing the location information exposed to any single party?

Contribution — Designed the full system architecture and implemented the entire Rust workspace: the gossip protocol, geo-topic manager, rendezvous service, centralized-broker baseline, mobility simulation harness, monitoring dashboard, and a real navigation website built on the same protocol.

Main result — Evaluation in progress. The Rust workspace and a real navigation website (self-hosted routing for Basel and Tirana) are built and running; the four research questions (dissemination feasibility, traffic and routing usefulness, privacy-utility trade-off, and robustness against false reports) have not yet been formally evaluated.

Research question — How should information systems be designed to remain consistent and available under distribution?

Contribution — Coursework covering distributed systems design principles and trade-offs.

Main result — Worked through the CAP-theorem trade-offs, replication and consistency models, and consensus protocols in a series of exercises and a group workshop applying these principles to a distributed system design.

Research question — How can cloud-native messaging services be used to build reliable, decoupled application components?

Contribution — Designed and implemented messaging-based application components on AWS.

Main result — Built decoupled application components using AWS messaging primitives (queues and pub/sub), covering delivery guarantees, retry/backoff behavior and failure isolation between producers and consumers.

Research question — What insights can graph-based modeling of flight-route data reveal that relational models make harder to see?

Contribution — Modeled flight-route data as a graph in Neo4j and analyzed connectivity and routing patterns.

Main result — Modeled airports and routes as a property graph and used Cypher queries (shortest path, centrality) to surface hub airports and connectivity patterns that were harder to see in the equivalent relational representation.

Research question — How do different continuous optimization algorithms behave across convex and non-convex objectives?

Contribution — Implemented and compared continuous optimization algorithms as part of coursework.

Main result — Implemented first-order and quasi-Newton methods from scratch in NumPy and compared convergence behavior across convex and non-convex objectives; published as a public repo with benchmark visualizations and an interactive Streamlit dashboard.