senior applied ai developer · sanovinci inc. · montréal

Agentic AI, from prototype to production.

I'm Hassan Mansour, a Senior Applied AI Developer contracting through SanoVinci Inc. I take agentic AI from prototype to production inside large, regulated organizations, in English or French: 6+ years building production systems at Bell Canada, and today at a major financial institution.

three ways to work together

Pick the format that fits.

A structured sprint for your organization, a workshop or talk for your team, or 1:1 sessions if you're learning to build.

for businesses

AI Adoption Sprint

A fixed-scope, 4–6 week sprint of working sessions that takes one AI use case from discovery to a working pilot your team owns. Any industry.

How the sprint runs

  1. Discovery
  2. Build with evals & guardrails
  3. Your team owns the pilot
  • Fixed scope: 4–6 weeks of working sessions
  • Any industry, early in AI adoption
  • Delivered through SanoVinci Inc., custom quote

Running a local service business? See ready-made automations.

See the AI Sprint
for teams

Workshops & talks

AI literacy for your team: a hands-on workshop for non-technical and mixed teams, in 2 hours or 3.5 hours. Plus talks and webinars for events and communities.

How the workshop runs

  1. Demos on your real tasks
  2. Hands-on prompting
  3. Your first 3 use cases
  • 2 hours or 3.5 hours, in person or remote
  • For non-technical and mixed teams
  • Prompt guide, use-case worksheet, risk checklist

Mixed business team already using AI? See AI at Work, Measured.

See workshops
for learners

For developers & students

You've watched the tutorials. Now you want to actually build with AI — and understand what you're doing.

One example, from many ways we can work: a 1:1 session

  1. Your project
  2. 1:1 session on your code
  3. Shipped, with notes
  • Learn to implement AI, not just talk about it
  • Project guidance from idea to shipped
  • Honest feedback — no fluff, no gatekeeping
Book a session

why trust me

I don't just advise. I ship.

Most AI advice comes from people who have never put a system in front of real users. My advice comes from the trenches.

  • 6+ years at Bell CanadaBuilding production systems, the last ~3 focused on AI. Now an AI developer at a major financial institution.
  • 2× Excellence and Innovation AwardBell Canada, 2026, for AI innovation and AI security · Bell Hackathon winner
  • NSERC & FRQNT scholarResearch scholarship recipient
  • MSc, McGill UniversityMultilingual Digital Communication (in progress)

Production AI is my day job

Today I work as an AI developer at a major financial institution. My playbooks reflect what holds up in production, not what trends on social media.

I build and run my own products

Maktab, my education SaaS, is live with real students. Ledger, my GST/QST assistant for Québec consultants, is in pilot with real users. I build them end-to-end, from architecture and AI features to every unglamorous part in between. When I recommend something, it's because I've lived it.

part of my toolkit

A sample, not the full list: the stack I use to build AI systems, and the AI tools I build with.

building ai systems
  • LLMs
  • AI agents
  • Multi-agent orchestration
  • LangGraph
  • MCP
  • RAG
  • Embeddings & vector search
quality & production
  • Evals
  • Observability & tracing
  • Structured output
  • AI security
  • Cost optimization
engineering & ai tooling
  • Python
  • FastAPI
  • PostgreSQL
  • pgvector
  • AI tooling

+ and more

selected work

What I have built, and where each one stands

Products from my own work, each built end-to-end. Every one carries its real status: live, coming at launch, or planned.

live · launched oct 2026

Maktab

An education platform for small organizations and private schools. I build it, run it and support its users.

It is live, with real students using it.

what it is
A learning platform for small organizations
built for
Small organizations and private schools
my part
Architecture, AI features, operations and support
status
Live since October 2026
in pilot with real users

Ledger

A GST/QST back-office AI assistant for incorporated Québec consultants. Receipt extraction with Claude vision, AI categorization, deterministic reconciliation, built with Québec's Law 25 privacy rules in mind.

It's in pilot with real users. If you're an incorporated consultant in Québec and want to be one of them, ask for pilot access.

live · open source

Proofline

An open-source, multi-agent content pipeline built with LangGraph. Nothing ships without passing the evals and a human sign-off.

Built in public under the Apache-2.0 licence. Everything in the repository is demo configuration and sample data, and no score is published: the model-graded evaluations have not been run yet.

Open the Proofline repository on GitHub

the measured family

One promise: every claim was measured.

Measured is the name my book, courses, workshop and tools share. Here is each one, with where it stands today. Nothing below is for sale on this site.

coming at launch

AI Engineering, Measured

The book. From Developer to AI Engineer, Backed by Reproducible Runs. It holds 143 measured findings, and every number in it came from a run. There is no release date and no order page yet.

Join the launch list
coming at launch

Measured Academy

Courses with a test in every module, for learners aged 18 and over. Course 1 follows the book, one module per chapter, and its first module is free once you have created a free account (sign-up required, no payment). No course is open yet: there is no enrolment and no price.

Planned courses: AI Vision, Measured (a working title) · Vibe Code a Real App · AI in Your Daily Work.

live

AI at Work, Measured

A hands-on workshop for mixed, non-technical business teams: 2 hours or 3.5 hours, for groups of 8 to 20. It is new, and the first pilot has not run yet, so there are no results to show.

See the workshop
planned

Measured Lab

A playground for checking AI outputs. A first version is built with one tab, Evaluate, which runs in your browser and calls no model. It is not launched. The Observe and Secure tabs are not built.

Cover of AI Engineering, Measured: From Developer to AI Engineer, Backed by Reproducible Runs. Hassan Mansour.

the book · coming at launch

AI Engineering, Measured

From Developer to AI Engineer, Backed by Reproducible Runs

A field guide for developers moving into AI engineering, with 143 measured findings. Fifteen chapters on agents, RAG, vector storage, multi-agent systems, MCP, evaluation, security, observability, cost and streaming, each with the tradeoffs, an architecture diagram, edge cases and a hands-on exercise.

Every number in this book came from a run.

how working with me goes

Simple, honest process. No lock-in.

Talk

A free 20-minute fit call. You tell me your situation; I tell you honestly whether I can help.

Plan

We define one specific outcome — not a vague "AI transformation." Something we can point to and measure.

Build

I implement, or I teach you to. Small first, working fast, expanded only when it earns it.

Own it

You end up with something that runs without me — and the understanding to keep it running.

latest writing

Notes from building

Read all posts

Let's figure out your next step.

Whether you run a business or you're learning to build — the first conversation is free and genuinely useful either way.