Bruno Kacper Mlodozeniec

AI Resident at Microsoft Research Cambridge. Formerly an Information and Computer Engineering (MEng) student at University of Cambridge. I have previously worked on uncertainty in deep learning, unsupervised learning, probabilistic modelling and causality.

I am passionate about unsupervised learning, principled decision-making, causality and representation learning, and other things that make non-iid settings exciting.


University of Cambridge

M.Eng. in Information Engineering

2016 - 2020
  • Master thesis: "Causal Inference: a Probabilistic Modelling Perspective", supervised by Professor Richard Turner.
  • Achieved first class performance in all years.
  • Ranked top 10%, 8%, and 7% of the year each year respectively (last year unranked due to COVID-19).
  • Co-founded and chaired the Cambridge University Artificial Intelligence society.


AI Resident

Microsoft Research Cambridge

09/2020 - 09/2021
  • Project 1: Bayesian Optimization and probabilistic modelling for synthetic biology.
  • Project 2: Reinforcement Learning for Compiler Optimisation.

Apple Software Engineering Intern

Siri Team

06/2019 - 09/2019
  • Worked as an ML Engineer on the Siri Voice Assistant.
  • Worked on unsupervised learning methods for distributional drift deteciton.

Cambridge Undergraduate Research

Machine Intelligence Laboratory

08/2018 - 10/2018
  • Worked with Professor Gales’ group on automated assessment ofr elevance of a spoken response to a given prompt in the context of language learning assessment.
  • Investigated various methods for training neural networks capable of providing useful measures of uncertainty, including deep ensembles and deep Prior Networks.
  • Investigated Ensemble Distribution Distillation using Prior Networks as a novel method for ensemble distillation that retains ensemble’s performance on uncertainty-related tasks.
  • See the paper: Ensemble Distribution Distillation.

Harvard SEAS Internship

Connectomics Research

07/2018 - 08/2018
  • Worked on developing methods for statistical analysis of networks representing the brain connectome.
  • Prototyped deep learning models for the purpose of identyfing repeating structural components within the brain network (graph motifs).
  • VCG's Connectomics Project Website.

Cisco Machine Learning Internship

Cisco Webex Teams

07/2017 - 08/2017
  • Worked on wake-word detection for the Cisco Webex range of products.
  • Trained and applied multiple deep learning architectures to the problem of keyword spotting.