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.

Selected work

DiscoverAI with Deloitte · 2026

Review-aware hybrid search over 60K Amazon products — RRF-fused dense + BM25 retrieval in Qdrant, cross-encoder reranking, and schema-enforced LLM summarisation.

P@5 0.952 · hybrid retrieval nDCG@10 0.951 schema validity 100%
AlgoWatch with Accenture · 2026

A Gemini-based agent that audits public-sector algorithms against the EU AI Act — ingest, evaluate, explain, remediate.

2nd of ~150 projects ~1,500 participants
Stablecoins vs. SWIFT 2026

Can stablecoins replace correspondent banking? Four hypotheses over 2020–2025, using the FTX collapse as a structural break — and three came out against the prediction.

log-log OLS · Newey–West two-way fixed effects 123-country panel
Thirsty Machines 2026

The energy, water and carbon cost of a single LLM query, mapped across data centres and grids — with a 184× spread between the lightest and heaviest models.

66 models · 37 regions 6 Tableau dashboards D3.js
CVE-2025-23211 individual · 2026

Reproducing a Jinja2 sandbox escape end to end — template injection to root RCE in a Dockerised target, then verifying the patch rejects the same input.

CVSS 9.9 exploit + fix demonstrated
Synaptic Dialogues individual · 2026

A week of my own AI usage, self-tracked and rendered as a navigable 3D spatial topography rather than a bar chart.

Three.js · WebGL interactive
Kalib personal · 2026

Built in my own time: a local-first nutrition PWA that measures maintenance calories from your own weigh-ins and logged intake instead of assuming a formula — with the uncertainty on the number shown rather than hidden.

offline USDA db · 13k foods measured TDEE · 95% interval 226 unit tests

How I work

Evaluation before enthusiasm

A number without a baseline is decoration. Every system gets compared against the simpler thing it was meant 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 the headline metric, and the faithfulness audit came in under the threshold I 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

Languages
PythonRSQLJavaScript
ML & AI
PyTorchTransformersvLLMscikit-learnXGBoostQdrant
Data & viz
pandasNumPyTableauPower BID3.jsThree.js
Infra
DockerPostgreSQLGitStreamlitKNIMEVercel

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