How we work
A small studio with a written method: measure first, build in small steps, report what fails, hand over everything.
Five steps
You can stop after any of them.
- 01
Samples or a brief
Send up to 3 sample documents, or describe one process in a few lines. Remove sensitive data first.
- 02
Measured feasibility (free)
We test on your material and tell you what is feasible, how we would measure success and the main risks.
- 03
Written proposal
Scope, deliverables and delivery date in writing. Changes are agreed in writing before extra work.
- 04
Build and test
Small steps, automated tests, results measured on your data. Uncertain outputs are flagged.
- 05
Handover
Source code you own, a plain-language guide and a short call. We install the tool on your infrastructure if needed.
How we measure
- Separate sets. A development set to build on, a hold-out set kept apart for the result, and a stress set of harder cases to find the limits.
- Counts with denominators. “28 of 28”, not “100%”; every number with its date and data.
- A pass mark agreed in advance, before the final measurement.
- Failures reported, including what was wrong without being flagged.
How we use AI
AI tools help with drafting, coding, testing and checking; Foysal reviews and is responsible for every result. In your project, AI steps are off by default and run only on your own API key or a local model, never on our subscriptions. AI-filled results are flagged for checking.
Tools we use
Python
All demos and research code
Streamlit
Internal apps and review screens
pandas
Data cleanup and checks
RapidOCR (ONNX Runtime)
Offline text recognition for scans
pytest
Automated tests on every build
Docker
Packaging so tools run the same everywhere
Optional AI
AI tools help with drafting, coding, testing and checking; Foysal reviews and is responsible for every result. In your project, AI steps are off by default and run only on your own API key or a local model, never on our subscriptions. AI-filled results are flagged for checking.
Only tools used in our own projects; each project page lists its full stack.
Who it is for
The kinds of organisations whose problems match what we build. Not a list of clients: we have none to show yet.
Accounting and bookkeeping
Invoices, receipts and bank documents retyped into spreadsheets every month.
Small trade and manufacturing
Supplier documents, delivery notes and order confirmations arriving as PDFs.
Research groups
Models to evaluate and code that should be reproducible.
Non-profits
Forms, applications and records handled by small teams with little time.
What we refuse
24/7 support, hosting or service-level agreements
Regulatory sign-off
Ghost-written academic work
Security audits
AI that would run on our own subscriptions
Project essentials
Is Snello one person?
Yes. Snello is one founder, Foysal, based in Italy, working with AI tools. Foysal reviews and is responsible for everything delivered.
Where is my data processed?
On your own computer where possible, or on ours for the feasibility check and the build. Documents are not sent to any AI service unless you choose one, under your own account. Sample documents are deleted after the feasibility check.
Do you use my documents to train AI?
No.
Who owns the code?
You do. Every project ships with source code and documentation.
Can you guarantee accuracy?
No. We measure accuracy on your samples and report it, including failures. Anything uncertain is flagged rather than guessed.
What languages do you work in?
English and Italian.
Bring your workflow into focus.
Send up to 3 samples or describe one process. We reply by email with what is feasible and how we would measure it.