my name is Emanuele Cannizzaro
I'm an engineer and CTO. I run the technical side of AECS4U, where we're turning Italy's judicial real estate auction market โ historically buried in scanned PDF court documents โ into structured, queryable data.
Before AECS4U I spent 25+ years in aerospace engineering (Airbus, Atkins, Ricardo, Assystem) on aircraft structural assembly and certification, while working in Python, ML and (more recently) Rust on the side. The PropTech pivot started when I noticed Italian judicial auctions have every hallmark of an inefficient market: the data exists but is locked in unstructured formats, valuation is artisanal, and capital allocation is dominated by local insiders. Worth fixing.
A document-understanding pipeline for Italian judicial real estate auctions. End-to-end:
- OCR & extraction โ olmOCR2 โ custom chunker โ Claude Haiku labeler โ ShareGPT JSONL training data
- Fine-tuning โ QLoRA on Qwen 2.5 via Axolotl
- Inference โ vLLM with
guided_jsonfor schema-constrained output - API layer โ FastAPI + SQLModel + PostgreSQL on DigitalOcean
Document types in scope: perizie tecniche, ordinanze and avvisi di vendita, contratti di locazione, visure storiche / attuali, and ispezioni ipotecarie.
A multi-source valuation surface for distressed property assets. Pulls from Idealista, Borsino Pro, Openapi SpA, OMI and CASAFARI into a confidence-weighted comparables view. Goal: replace artisanal valuation with something an investor can underwrite from.
A three-component ATS scoring system (CV / cover letter / job-description
match) with a 17-check ATS engine and a multi-profile system that swaps
between career tracks โ CTO, senior engineer, PropTech / data, aerospace /
systems, EU institutional โ via a profiles.json selector.
Aerospace โ 25+ years across Airbus, Atkins, Ricardo and Assystem. Most recently Technical Team Leader on the A350 nose and centre fuselage structural assembly team. Chartered Engineer (CEng).
Software โ Python since the late 2000s, Rust since 2022, TypeScript when something has to render in a browser. FastAPI, SQLModel and PostgreSQL are my default stack for shipping production services.
ML / NLP โ Master's in Non-Performing Loans Management (2025). Active in fine-tuning open-weight models for legal-domain document understanding and structured extraction.
GIS โ geospatial work threads through everything I do at AECS4U: multi-floor cadastral mapping, competitive analysis of ~70 Italian auction platforms, comparative studies of judicial-auction transparency across 25+ countries. I publish technical material in this space on LinkedIn.
You can reach me at
