Skip to content
View Giotor84's full-sized avatar

Block or report Giotor84

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Giotor84/README.md

Gioacchino Tortorici — AI su documenti legali e finanziari in ambienti chiusi

Porto l'AI generativa dove i dati non possono uscire: servicer NPL, SPV, banche e studi di recupero crediti. Costruisco l'intera catena (acquisizione dei documenti, OCR, estrazione con modelli locali o in cloud privato, verifica, dashboard) e misuro quanto funziona, anche quando non funziona.

Come iniziare

passo cosa durata che cosa ottenete
Diagnostica un piccolo campione dei vostri documenti, per esempio 10 fascicoli 2 settimane errori, casi "da verificare" e tempi misurati, senza impegno a proseguire
Pilota un campione più ampio, con un campione di confronto fissato prima di far girare il sistema 6–8 settimane dall'accesso all'ambiente, secondo il prodotto report di accuratezza, che mostra anche gli errori, e decisione
Licenza annuale prodotto e supporto annuale il costo del pilota è scalato dal primo canone se firmate entro 6 mesi

Tariffa fissa concordata prima dell'avvio. Il lavoro avviene nella vostra infrastruttura o nel vostro account AWS senza internet, sotto NDA.

Contatto: gioacchino.tortorici@gmail.com. Indicate azienda, caso d'uso e volumi indicativi; nessun dato riservato nel primo messaggio.


Commercial products (on-premises & air-gapped cloud, available under NDA)

product buyer & problem solved deployment
ZT Legal Analysis NPL servicers, banks, SPVs, recovery law firms: turns credit-recovery dossiers of hundreds of PDFs (native + scanned deeds) into a draft legal report and actionability checklist for lawyer review; amounts, prescription deadlines and chain-of-title guardrails computed in Python Local GPU or air-gapped AWS (SageMaker + Bedrock)
ZT Orienta Litigation heads & NPL portfolio managers: extracts structured rulings, norms and debtor exceptions from your private corpus of Italian court decisions, with human review gate and offline forum/section orientation dashboards Air-gapped AWS or private cloud + offline HTML/Excel
ZT Call Analytics Debt-collection call centers & servicers: transcribes collection calls with speaker separation, flags promises to pay (amount + ISO date) and compliance issues with verbatim quotes for supervisor review; accuracy measured on your own sample during the pilot Air-gapped SageMaker + Bedrock or on-prem GPU

Public repositories

project what license
zt-batch-reader Open-core closed document extraction via calibrated first-token logprobs, abstention & deterministic rules (commercial 284B MoE GPU streaming add-on available) PolyForm Small Business + commercial
zt-anchor Document AI that can say "I don't know": verbatim anchoring, abstention, blind-window backtest, frozen-holdout evaluation, Italian contract lint (live demo) PolyForm Small Business + commercial
zt-airgap Tested template for LLM pipelines on SageMaker with no internet: verified releases via S3, Bedrock via proxy, numbers in Python Apache-2.0
zt-legal-analysis-core Source-available core of ZT Legal Analysis: OCR cascade, resumable checkpoints, rolling synthesis for case files larger than the context, YAML decision rules, LLM adapters PolyForm Small Business + commercial
zt-legal-graph Components of a local pipeline for drafting creditor-side replies in Italian civil cases, with claims anchored to a statute graph built from Normattiva PolyForm Small Business + commercial
zt-memory Local code-context engine + MCP server: makes local and air-gapped LLMs useful on real codebases (right file in context 93–100% vs 21–29% with whole files) Apache-2.0

What I have built in production (client work, described without confidential details)

  • Legal intelligence on case law: thousands of Italian court decisions turned into a searchable graph of rulings, statutes and exceptions; extraction with a private-cloud LLM, human approval before publication, offline dashboards for executives and lawyers.
  • From case file to legal report: OCR cascade and per-document-type extraction over files of hundreds of PDFs (mortgages, guarantees, enforcement deeds), producing a narrative report and an action checklist.
  • Call analytics for collections: on-device transcription with anti-hallucination filters, compliance and promise-to-pay scoring, KPI dashboards; plus an offline real-time call translator (no audio leaves the PC).
  • Document acquisition and automation: supervised document acquisition, legacy terminal automation, public-auction extraction with code-side validation.
  • Portfolio analytics: recovery-rate and collections forecasting with blind-window backtests, segmentation by next best action, settlement NPV tools.

Stack: Python · llama.cpp / GGUF · Amazon Bedrock & SageMaker (air-gapped) · Ollama · Whisper · tree-sitter · SQLite FTS5 · Playwright · MCP.


Work with me

  • Diagnostic and pilot on your own documents or call recordings, at a fixed fee agreed upfront: inside your infrastructure or air-gapped AWS account, under NDA, with the evaluation sample and success criteria fixed before the system runs. The pilot fee is credited toward the first annual license if signed within 6 months.
  • Commercial annual licenses (at or above the PolyForm Small Business threshold of 100 people or 1M USD revenue) for zt-batch-reader, zt-anchor, zt-legal-analysis-core and zt-legal-graph, with a free 90-day internal evaluation.
  • Contact: gioacchino.tortorici@gmail.com. Please indicate company, use case and approximate volume (never send confidential data in the first message).

Popular repositories Loading

  1. Giotor84 Giotor84 Public

    Applied AI systems for high-stakes legal & financial documents: local GGUF inference, anchoring & air-gapped pipelines

  2. zt-memory zt-memory Public

    Local code-context engine + MCP server that feeds LLMs exact spans instead of whole files

    Python

  3. zt-airgap zt-airgap Public

    Tested template for air-gapped LLM pipelines on AWS SageMaker & Bedrock with zero internet

    Python

  4. zt-anchor zt-anchor Public

    Stop LLM hallucinations on documents: verbatim quote verification, abstention, holdout eval & blind backtesting

    Python

  5. zt-legal-analysis-core zt-legal-analysis-core Public

    Process oversized legal PDF dossiers locally: OCR cascade, rolling synthesis, YAML decision rules & air-gapped LLMs

    Python

  6. zt-legal-graph zt-legal-graph Public

    Local, source-anchored legal drafting for Italian civil cases using an official Normattiva law graph

    Python