AI consulting for retail and CPG teams¶
I’m Franco Betteo, an AI engineer and senior data scientist. I help teams turn a specific business problem into an AI or machine learning system they can evaluate, use, and maintain.
My work spans demand forecasting, optimization, pricing analysis, segmentation, product discovery, and applied AI. I have worked at Nielsen, InsiteAI, and Despegar, and with teams ranging from startups to the University of Oxford.
See selected projects Explore consulting services
Where I can help¶
- Find a worthwhile use case. Assess the data, expected value, risks, and a practical first experiment.
- Build and evaluate a prototype. Work with your team to test whether an approach solves the actual problem before planning production work.
- Improve an existing system. Review model quality, data pipelines, evaluation, and the path to reliable use.
I work directly with your team and explain the tradeoffs behind the solution. The goal is a system your team can understand and improve.
Selected work¶
Much of my consulting work is for private clients, so I cannot share their data, code, or a public demo. Here are the kinds of decisions and workflows I have worked on:
- Demand forecasting: predicting future demand with gradient-boosted models, including LightGBM, so teams can plan with a view of likely demand.
- Optimization: using linear optimization with PuLP to compare feasible choices under business constraints.
- Pricing and segmentation: analyzing price elasticity and grouping customers or products with methods such as k-means and DBSCAN to support more targeted decisions.
- Visual product discovery: using image embeddings to find similar items in an ecommerce catalog.
The projects page explains what I can share and how these methods relate to a buyer's problem.
Product discovery for an ecommerce catalog¶
I built an image similarity system to help a retailer find related products in its catalog. The work included image collection, visual embeddings, a FAISS index, and an API for retrieving similar items. The project page explains the scope, implementation, and what is known about its use.
Read the project · Try the portfolio demo
A public product I build¶
Download YouTube Transcripts is a product I build and operate for extracting available transcripts from individual videos, channels, and playlists. It offers bulk processing and an API, including for teams preparing text for AI workflows.
What clients have said¶
“The educational materials Franco developed for our ML program were exceptional. A complex task successfully completed.”
— Ajit Jaokar, Fellow, Department of Engineering Science, University of Oxford
“Franco developed the first MVP using Probabilistic Graphical Models to solve the main goal of Diggity.”
— Stephan, CEO, Diggity "Working with Franco on our forecasting models gave us a competitive edge in our market."
— Juan Bujacich, Manager, Nielsen
Discuss a project¶
If you have a retail, CPG, or applied AI problem, send me a short description of the decision or workflow you want to improve, the data you have, and your timeline. I’ll tell you whether I can help and suggest a sensible first step.
Book a 30-minute introduction or email francobetteo@gmail.com.
I also write about statistics, machine learning, and building software. Some articles are in Spanish.