Nico Schiavone

Currently, I work as a software engineer at Microsoft.

I am an M.Sc. graduate in CS from the University of Toronto and a B.Sc. graduate in Engineering Physics from the University of Alberta. I was fortunate to be advised by Sheila McIlraith, Eldan Cohen, Xingyu Li, Frank Marsiglio, and Tian Tang.

In the past, I was researching multi-agent systems, game theory, neutrinos and superconductivity. I enjoy drawing and writing in my free time.

1p resume  /  email  /  github  /  scholar  /  linkedin

/ uniqlo /  substack

Research

Contracts Improve LLM Cooperation

We show how structured contracts with a negotiation phase in mixed-motive games improve the performance of LLM agents, closing the gap between non-thinking and thinking models, and greatly adding to the power of multi-agent systems made from lower strength models.

It's Rational for AI Agents to Procrastinate

We encode laborious chores in a new type of game, the labour game, and analyze how agents behave when they are disincentivized to contribute but incentivized towards completion. This closely mimics scenarios like group projects, house cleaning, and other human-human tasks. We find the emergence of a phenomenon like procrastination, and define rules that govern an agent’s behaviour in this game.

Reinforcement Learning with Generative Models for Compact Support Sets

In this work, we propose a framework utilizing reinforcement learning as a control for foundation models, allowing for the granular generation of small, focused synthetic support sets to augment the performance of neural network models on real data classification tasks.

MyriadAL: Active Few Shot Learning for Histopathology

Oral Presentation (Top ~5% of Accepted Papers). In this study, we address this issue in the scenario where the annotation budget is very limited, yet a large amount of unlabelled data for the target task is available. We frame this work in the context of histopathology where labelling is prohibitively expensive.

Other Projects

Nutrition

A tool to find out what nutrients you are missing from a plaintext, natural language description of your diet. Also suggests foods to cover missing micronutrients. Uses nutrition data from the FoodData Central database and Streamlit for a clean visualization.

UNIQLO Web Scraper

A web scraping project using Scrapy and Playwright, developing into a command-line tool for easy shopping on ecommerce sites. Check it out here.

HUGINN

Autonomous retrieval drone made for a capstone project. Utilizes ROS and YOLOV8 to autonomously find the object of choice, and magnetic grippers to pick it up.

Ubiquitous researcher template from Jon Barron's website