Eliya Allam

Data Science & Management · LUISS Guido Carli, Rome

I build retrieval and machine-learning systems — and I publish the numbers, including the ones that don't flatter me.

MSc candidate in Data Science & Management, following a BSc in Management & Computer Science graded 107/110. Most of what I build is a pipeline of some kind — retrieval, inference, econometrics — and the part I care about is whether the result survives scrutiny: an evaluation you can trust, limitations stated before someone else finds them, and a repo another person can actually run.

Currently open to internships and job roles alongside study.

Selected work

DiscoverAI with Deloitte · 2026	Review-aware hybrid search over 60K Amazon products. RRF-fused dense + BM25 retrieval in Qdrant, cross-encoder reranking, schema-enforced LLM summarisation. P@5 = 0.952 · nDCG@10 = 0.951 · 100% schema validity
AlgoWatch with Accenture · 2026	A Gemini-based agent auditing public-sector algorithms against the EU AI Act — ingest, evaluate, explain, remediate. 2nd place of ~150 projects · ~1,500 participants
Internet & Network Economics 2026	Can stablecoins replace SWIFT? Four hypotheses over 2020–2025, using the FTX collapse as a structural break. Log-log OLS with Newey–West errors · two-way fixed-effects panel regression
Thirsty Machines 2026	The energy, water and carbon cost of a single LLM query, mapped across data centres and grids. 66 models · 37 cloud regions · 6 Tableau dashboards + D3.js
CVE-2025-23211 2026	Reproducing a Jinja2 sandbox escape end to end — template injection to root RCE, then verifying the patch. CVSS 9.9 · Dockerised · exploit and fix both demonstrated
Synaptic Dialogues 2026	A week of my own AI usage, self-tracked and rendered as a navigable spatial topography rather than a bar chart. Three.js · deployed

How I work

Evaluation before enthusiasm. A number without a baseline is decoration. Every system I build gets compared against the simpler thing it was supposed to beat — sometimes the simpler thing wins, and that is worth knowing early.

Limitations in the abstract, not the appendix. DiscoverAI's cross-encoder reranking lowers our headline metric, and the faithfulness audit came in under the threshold we set in advance. Both are in the first paragraph of the report. Work that hides its weak points is harder to trust than work that names them.

Reproducible or it didn't happen. Pinned model SHAs, fixed seeds, per-stage validation gates. Same seed and same hardware, same artefacts — otherwise a result is an anecdote.

Toolkit

Education

MSc Data Science & Management — LUISS Guido Carli, Rome · 2025 – 2027
BSc Management & Computer Science — LUISS Guido Carli, Rome · 2022 – 2025 · 107/110
_{Thesis: The Impact of Artificial Intelligence on OSINT Technologies}

_{Cisco Cybersecurity · Celonis Build Analyses · LUISS AI Literacy}

_{Rome · English, Russian and Arabic fluently; Italian conversationally.}