AI software that works the way your business does
We build custom machine learning pipelines, prediction engines, and automation tools for companies across Quebec and beyond.
We build custom machine learning pipelines, prediction engines, and automation tools for companies across Quebec and beyond.
Forward AI Solutions started in 2021 when two machine learning engineers from Université de Sherbrooke decided to stop building models for research papers and start building them for real businesses. We got tired of watching local manufacturers, retailers, and service companies struggle with problems that a well-trained model could solve in seconds.
Our team now has six people. Two focus on data engineering, two on model development, one handles project management, and one writes the documentation nobody else wants to write. We operate from 245 Rue Frontenac, Sherbrooke, Quebec J1H 1A1, Canada.
Most of our clients are mid-sized companies with 20 to 500 employees. They have data sitting in spreadsheets, ERPs, or databases, and they want that data to actually do something useful. That is where we come in.
Every project starts with your data and your problem. We do not sell off-the-shelf products. Here is what we typically deliver:
We train models on your historical data to forecast demand, churn, equipment failure, or whatever metric drives your decisions. A recent client in food distribution cut spoilage by 23% using a demand forecasting model we built on 18 months of their sales records.
Customer emails, support tickets, product reviews, contracts: we build classifiers and extraction pipelines that read these documents and pull out the information you need. One insurance broker uses our system to extract claim details from freeform emails and route them to the right adjuster in under two seconds.
Quality inspection on production lines, inventory counting from shelf photos, document digitization. We have deployed vision models for a furniture manufacturer that detects surface defects on wood panels before they reach the finishing stage. The system catches defects the human eye misses at 4 AM.
Before a model can learn anything, your data needs to be clean, connected, and flowing. We design ETL pipelines that pull from your existing systems, whether that is PostgreSQL, Salesforce, SAP, or a folder of CSVs someone emails every Monday. We use Apache Airflow and dbt for orchestration.
Not the generic chatbot that annoys your website visitors. We build retrieval-augmented generation (RAG) systems trained on your own knowledge base, product catalog, or internal documentation. Your staff or customers ask a question in plain language and get an answer grounded in your actual data.
A model is useless if it sits on a laptop. We package our models as REST APIs and integrate them into your existing software stack. We handle deployment on AWS, Azure, or on-premise servers depending on your security requirements and budget.
We follow a structured process, but we keep it flexible enough to adapt when data surprises us. Here is the typical flow:
Real work happens in terminals, not in boardrooms.
Small projects (a single prediction model with API) start around $15,000 CAD. Larger engagements with multiple models, data pipeline work, and integration can run $40,000 to $120,000. We give you a fixed quote after the data audit, not before.
Yes, but we sign an NDA before we see anything. For clients with strict data residency rules, we can work entirely on your infrastructure. We have done projects where our developers never downloaded a single file; everything ran inside the client's VPN.
It depends on the problem. Some tasks need thousands of examples; others work well with a few hundred. During the data audit we will tell you honestly whether your dataset is large enough. If it is not, we can sometimes use transfer learning or synthetic data generation to bridge the gap.
Python is our primary language. We use PyTorch and scikit-learn for modeling, FastAPI for serving, and Docker plus Kubernetes for deployment. For data pipelines, we rely on Apache Airflow and dbt. If your stack requires something specific (Java, .NET), we can adapt.
Absolutely. About half our projects involve close collaboration with an in-house developer or data analyst. We use Git for version control and hold weekly standups over video call. Your team gets full access to the code repository from day one.
We do. Most of our Quebec clients are in Sherbrooke, Montreal, and Quebec City, but we have worked with companies in Ontario and New Brunswick as well. Remote collaboration works fine for software projects. We visit on-site when needed for hardware integration or workshops.
Phone: +1 819 290-9128
Email: [email protected]
Address:
245 Rue Frontenac, Sherbrooke, Quebec J1H 1A1, Canada
Office hours: Monday to Friday, 9:00 AM to 5:00 PM Eastern.
We respond to emails within one business day. If your request is urgent, call us directly.