Eziyo EhsaniAI Solution Architect
Quantization, distillation and retrieval at the edge — working out what a language model actually costs in latency, energy and accuracy once it has to run privately, without the cloud.
01 — About
I design AI systems that solve real business problems and stay cheap to run. As an AI Functional Analyst at Sunprime in Milan, I turn business needs into system designs and solution architectures: which parts the system needs, how they connect, and where AI really adds value.
I look for the simplest design that works. Often that means a smaller or quantized model instead of a large one, caching and batching instead of more servers, and running models on the device or on premises when the data allows it. My research on on-device LLMs taught me to measure every choice in latency, energy and money before it reaches the bill.
Before anything goes to production, I test the idea quickly. I build a small working prototype, try it on real data, and use the results to decide: build it, change it, or drop it. SolarFlow, the solar pipeline dashboard on this site, is one example.
A prototype is only useful if it can grow. I review code, write tests for the parts that matter, and check AI output against clear evaluation sets, so the team can trust the system it ships.
My background is an MSc in Data Science from Federico II and five years in data science and ML engineering.
Thesis · Green Mobile Intelligence
Quantifying energy–accuracy trade-offs in on-device LLMs across eight models, four quantization schemes and real mobile silicon.
On-device
0.5B–9B parameters, deployed with llama.cpp
Profiled
Isolated inference runs: latency, energy, accuracy
Modeled
Records behind churn & HR prediction models
MSc grade
Data Science, University of Naples Federico II
Experience
Across data science and ML engineering
Languages
Under review
Empirical Software Engineering
arXiv:2603.2660302 — Experience & Education
Five years across data science and ML engineering, now designing AI solution architectures at Sunprime.
Sunprime
Software and Sustainability (S2) Group, Amsterdam
DIETI Lab — University of Naples Federico II
Trained and evaluated transformer-based NLP models (BERT, DistilBERT) for text classification, comparing full-precision baselines against quantized on-device variants. Applied explainable AI techniques (SHAP, LIME) to clarify model behaviour and support accuracy–efficiency trade-off decisions, and built supervised learning pipelines in Python (scikit-learn, PyTorch) balancing latency, energy use and predictive performance.
HascoNet
Automated KPI dashboards in Power BI, cutting manual reporting time by 30%. Built predictive churn and HR analysis models over 100k+ records to inform retention and staffing decisions, and designed the ETL pipelines feeding analytics and machine learning workloads.
HascoNet
Built Power BI data models and DAX measures enabling dynamic filtering and performance monitoring, and supported senior analysts with reporting, data validation and documentation across analytics projects.
University of Naples Federico II
Thesis: Green Mobile Intelligence — Quantifying Energy-Accuracy Trade-offs in On-Device LLMs. Coursework in AI Systems Engineering, Data Mining & Machine Learning, and Hardware & Software for Big Data.
Azad University
Thesis: Design of an IoT-Based Energy Consumption Monitoring and Optimization System. Coursework in System Design & Analysis, Compiler Design, Microprocessors and Computer Architecture.
Quantifying energy–accuracy trade-offs in on-device LLMs.
How far compression can be pushed before quality collapses, measured across eight open-source models, four quantization schemes and 480 isolated inference runs on real mobile silicon.
03 — Selected work
Systems built end to end — from quantized inference on real handsets to fully offline retrieval and speech.
04 — Research
One manuscript under review, with the preprint openly available on arXiv.
2026
Under reviewEmpirical Software EngineeringEMSE · Springer
05 — Skills & Certifications
Methods, frameworks and tooling used across research, deployment and production data work.
06 — Resume
Experience, education, publications and the full tool list, in one PDF.
