Andrea CundariChartered accountant who builds AI systems · Italy (CET) · available for UAE remote work
I am a chartered accountant. I also build the systems that do my profession's work.
In Italy I am a Dottore Commercialista, in practice since 2007 and registered with the Order of Chartered Accountants of Cosenza: statutory accounts, tax litigation, business valuations under Italian PIV standards, and court appointments as insolvency trustee, expert witness and judicial administrator. In parallel, over eight weeks in 2026, I designed, built and released nine production software systems. Three are live right now. That combination is the whole point: I know which professional processes are worth automating because I have done them by hand, under professional liability.
9systems shipped
1,075commits
168,000lines of code
8weeks elapsed
Measured 16 Aug 2026 from the repositories themselves, across 25 Jun to 16 Aug 2026. Built with intensive AI-assisted development, which is the point: one operator, architecture and review kept human, delivery in weeks.
The professional practice
Why an accountant is the right person to automate professional work.
Most people who automate a professional firm have never worked inside one. They build what the brief describes, miss the exception that matters, and hand back a system that fails on the first non-standard file. I have sat on the other side of that.
Nineteen years of practice. Company accounts and tax, tax litigation before the Italian tax courts, business valuations under the Italian PIV standards with the full DCF and WACC workings reconcilable line by line, real-estate appraisals, and court appointments: insolvency trustee, expert witness, and judicial administrator of businesses under court-ordered administration, where the job is keeping a real company operating while a court supervises every decision.
That is the source of a rule I apply to every AI system I build for regulated work: the professional stays accountable, so the machine must never be able to pretend otherwise. No invented legislation, no invented case law, no citation that is not in the provided documents, no deadline presented as final, no gap quietly filled with an assumption. Not as a policy. As code, with tests that fail the build.
For clients in the Italy to UAE corridor this matters twice over: I can advise on the Italian side of moving a business or a residence abroad, and I can build the systems that carry the work. Italian professional practice is regulated in Italy, and I do not hold a UAE licence: I work alongside licensed local operators, never in place of them.
What I build
Four things, each with something built behind it.
Pull data out of documents that resist it
Scanned PDFs, image-only files, inconsistent layouts. Vision models with schema-constrained output, then automated checks that decide whether the result is trustworthy.
Proof: 100% accuracy on the tested flyer set
Run pipelines that need nobody
Scheduled ingestion, extraction, validation and storage, with failover between providers when one refuses or runs out of quota, and a record of which source produced every value.
Subscriptions, credit systems, paid APIs with keys, rate limits and signed webhooks. I design the pricing and the tiers, not only the endpoints.
Proof: 4 products with live payment processing
Make AI safe to put in front of a regulator
Hard prohibitions on inventing sources, mandatory attribution, fixed output contracts, and tests that fail the build when a guardrail is missing.
Proof: 7 encoded prohibitions, enforced by test
Selected systems
Every number below can be checked in a repository.
Live in production
Built
Jul to Aug 2026
Commits
816
Lines
56,603
Role
Sole builder
Real-time market reaction platform
A subscription product and a paid public API. It watches for public posts, measures how the monitored markets moved in the five minutes after each one, issues a verdict, and delivers the event to subscribers and to third-party clients.
The whole chain is automatic: detect, deduplicate, filter, activate, measure at T+5, close on recovery, judge, notify, distribute. Seven production cron jobs plus a wake-up channel.
Three market-data providers with controlled degradation. When the metered provider exhausts its quota the system falls back, and every measurement records which source produced it.
A symbol is not accepted because it responds, but because it responds for the same instrument. Cross-source verification catches the ones that quietly do not.
I designed the commercial side too: subscription tiers, an annual API plan with key and rate limits and three webhook endpoints, a delayed sandbox tier reusing the same code path, and contract-based redistribution.
Live in production
Built
Jul to Aug 2026
Commits
86
Accuracy
100% on test set
Role
Sole builder
Autonomous retail pricing pipeline
You paste a shopping list in plain language. It tells you, line by line, the cheapest matching offer in the valid flyers near you. The hard part is not the search, it is that the source documents are image-only PDFs with no text inside.
Runs unattended. The nightly cycle starts at 02:00 and finished ~1,500 offers across six retail chains in 60 minutes with no human involved.
Extraction uses vision models with JSON-schema output. Measured accuracy on the test set: 100% across ~42 products, including dense pages and non-food items.
It refuses to publish what it cannot verify. Unit prices are recomputed and compared against the price-per-kilo already printed on the flyer, handling packaging formats and multipacks. Anything that does not reconcile is quarantined for review.
Model failover proven live: when one vision provider rejected pages, a second recovered them. One generic ingestion framework with an adapter per chain, not single-use scrapers.
In service since Jul 2026
Providers
4
Lines
10,750
Role
Sole builder
Multi-provider market data API
A financial data API behind one interface, running on a hardened VPS. Honest status: it is in service and healthy, and it does not yet have external consumers.
Four providers behind a single contract, with credit rationing driven by a market clock: closed markets consume no quota, because wasted calls are wasted money.
Rootless Docker, bound to loopback only, never exposed directly to the internet.
Client keys are stored only as SHA-256 hashes. No secret ever enters version control.
Built
Built
Aug 2026
Lines
28,873
Domain
Legal and tax
Guardrailed AI for regulated professionals
An agent platform for lawyers and tax advisors, built around one non-negotiable rule: accountability stays with the human professional. Every output is a working draft subject to their verification, and the product is engineered so it cannot pretend otherwise.
Seven absolute prohibitions encoded as guardrails: never invent legislation or case law, never cite a source absent from the provided context, never guarantee an outcome, never present a deadline as final, never fill a gap with an assumption dressed as fact.
Every agent must add at least two guardrails specific to its own subject matter, and an automated test fails the build if it does not.
A fixed ten-section output contract, including missing data, stated assumptions, risks, and a verification checklist for the professional.
Three-label source attribution separating uploaded documents, the knowledge base, and everything still to be verified.
Built
Languages
EN, IT, AR
Lines
11,967
Market
UAE
Trilingual UAE pre-engagement platform
A platform that structures and qualifies European founders' UAE enquiries before they reach a licensed local operator. English, Italian and Arabic, with right-to-left support.
Compliance rules are enforced in the product, not in a policy document: no outcome promises, no fabricated metrics or testimonials, no company logos presented as clients.
Contact details are disclosed only through an explicit consent flow, and in demonstration mode they do not exist in the data at all.
The scope of what the service does not do is stated openly: it does not incorporate companies, issue licences, file visa applications or give advice.
Live with real traffic
Since
Jun 2026
Records
Millions
Stack
PHP 8, no deps
Public fraud lookup service
One question, one answer: is this phone number, website or email address a scam? It searches a database assembled from dozens of open feeds and returns a verdict.
A single SQLite database holding millions of records in WAL mode, with schema and indexes created and migrated on first access.
Plain PHP 8 with no framework and no dependencies, chosen deliberately so it runs reliably on shared hosting.
Deploy discipline: nothing overwrites production before the live file is downloaded and compared, and every release is verified by behaviour on the real site.
How I work
The method is the reason the systems survive contact with production.
Specify before building
Design happens in its own session and produces a written specification, reviewed before a line of code exists. Ambiguity is cheaper to fix in a document.
Build in a clean room
Implementation starts from the reviewed specification, with the open questions asked first, and commits at every intermediate step rather than one at the end.
Review independently, before deployment
A second, separate session reviews the diff with one job: find duplication, dead code, exposed secrets, regressions, logic holes. Every reported defect is verified against the real code before anything is changed, because reviewers are wrong sometimes too. On one product this produced sixteen findings across two passes, with no false positives. The author is the worst placed person to find their own holes, and a defect caught here never reaches the server.
Test by behaviour, not by reading
The question is always "when I do X, does Y happen", checked against the running system. On the credit-based products that meant 50 unit tests and 17 integration tests against a real PostgreSQL, covering atomic credit consumption, refund idempotency and recovery of jobs the queue never picked up.
Working together
Start small. The first step is deliberately easy to say yes to.
Automation assessment
EUR 1,500 · 3 days
I map the processes costing your team the most manual hours, say which are automatable and which are not, estimate the hours saved per month, and give you the three priorities in writing with what each would cost. No obligation beyond that.
First system in production
EUR 6,000 to 9,000 · 3 to 4 weeks
One system that does one thing, automatically, with quality gates that quarantine whatever it cannot verify. Deployed, tested by behaviour on the real thing, documented, and handed over.
Ongoing operation
EUR 1,500 to 3,000 per month
The system stays up, watched and maintained. Fixes, adaptation when upstream sources change, small evolutions, agreed availability.
I contract through Horizon Peak S.r.l.s., an EU-registered company, so a UAE client can engage me on a service contract without visa or employment overhead. That is also how UAE companies normally buy from Europe.
Contact
Tell me the process that is eating your team's hours. I will tell you honestly whether it can be automated.