PDF · 4 PAGES
AI scoping checklist
Twelve questions to ask before funding an AI project — and the answers that should raise a red flag.
ARTICLES
Engineering notes
What we learn putting systems into production: evaluation, costs, guardrails, operations.
LangChain and LangSmith: building and steering AI agents
LangChain is for building AI agents; LangSmith is for observing, evaluating and deploying them. This guide explains what each tool does, how they depend on each other, what they cost and where your data goes.
Read the article →
Node2Shell: the Brotli flaw that goes beyond Node.js
An error of a few bytes in Google's Brotli compression decoder exposes Node.js, Bun, Python and several cloud...
Your tools don't talk to each other: what an API changes for an SME
Website, cash register, accounting, mobile money, WhatsApp: most SMEs already have the right tools. What costs...
NGINX: five vulnerabilities fixed, three rated critical
On October 14, 2026, NGINX fixed five vulnerabilities, including three memory corruption flaws rated 9.2 out o...
An AI agent in production: what nobody tells you
The prototype works in three days. Going into production takes three months: evaluation, guardrails, costs, hu...
What AI changes in a Guinean SME
Beyond the hype, five uses actually adopted by SMEs in Conakry this year, and two that were dropped.
PHP2Shell: two critical flaws in the PHP engine
Disclosed on October 21, 2026, two flaws in the PHP engine allow remote code execution on a web server, withou...
Evaluating a model when data is missing
Three years of history, gaps, and a decision to make: how to evaluate a model honestly when the test set is to...
Name your tables the way you name your variables
A naming convention fits on one page and saves hours for every new person who joins the project.
RAG: the chunking mistakes that cost you dearly
The quality of a document search engine is decided before the model: in the way the documents are chunked.
A CI/CD pipeline that fits on one page
A pipeline the whole team can read is better than an exhaustive pipeline nobody understands. Here is the one w...
Cutting your cloud bill without breaking production
Three levers applied at a telecom client, without degrading the service: right-sizing, storage lifecycles and...
Our deep dives
DOSSIER 01
Taking an agent to production
From the prototype that impresses to the service you monitor: evaluation, guardrails, costs and human oversight.
Open →DOSSIER 02
Regaining control of infrastructure
Inventory, infrastructure as code, continuous deployment and monitoring, when nobody knows how it works anymore.
Open →DOSSIER 03
From data to decision
Naming, cleaning, modeling, presenting: the full chain from a messy database to a decision you can stand behind.
Open →Downloadable resources
PDF · 6 PAGES
Cloud selection matrix
Compare AWS, Azure and GCP on your own criteria: cost, sovereignty, available skills.
DOCX
Requirements template
A ready-to-fill outline for consulting several vendors on a comparable basis.
Who writes
PUT IT INTO PRACTICE
Relevant to your work?
We apply these lessons with our clients — and we teach them in our training courses.
+224614686869
contact@visiodev.net
Lambanyi, Conakry — Mon–Fri 8 am–6 pm GMT