Salvatore Esposito

Product-driven Engineer

Microsoft & University of Edinburgh

Biography

Product-focused engineer with a Ph.D. in Artificial Intelligence from the University of Edinburgh, specializing in 3D reconstruction and generative AI. Contributed to innovative AR/VR initiatives at Microsoft, including HoloLens and Teams Avatars projects, combining technical expertise with a keen understanding of user needs. Strong track record of collaborating across engineering, design, and business teams to deliver impactful solutions. Leverages research background and hands-on development experience to bridge technical feasibility with product vision. Dedicated to fostering collaborative environments that accelerate cutting-edge R&D and drive strategic business growth.

Interests

Technical Product Development
Systems Integration & Roadmapping
Generative AI & Computer Graphics
Data Engineering & Feature Pipelines

Education

PhD in Biomedical Artificial Intelligence, 2025
University of Edinburgh
MRes in Artificial Intelligence, 2021
University of Edinburgh

Professional Experience

Research & Development

Microsoft

January 2022 – January 2024 Zurich & Cambridge

Visiting Researcher, Zurich (2022-2024)

Led the enhancement of the HMR 2.0 pipeline, focusing on the development of a multimodal transformer for avatars
Specialized in predicting motion and audio, significantly improving the realism and responsiveness of avatar interactions

Research Intern, Cambridge (2022)

Developed and refined generative models, including GANs and neural radiance fields, focusing on enhanced surface representation quality
Innovated a novel generative model using a neural rendering architecture for improved extraction of surfaces from human facial meshes

Research Intern

Huawei

January 2023 – December 2023 Edinburgh

*Systems Infrastructure Research (2023)

Specialized in time series forecasting and predictive models for CPU utilization in VMs
Developed comprehensive forecasting solutions for system resource management

*Research Intern, Cambridge (2022)

Integrated neural radiance fields with caching mechanisms to enhance rendering quality

Quantitative Researcher

American Express

January 2022 – December 2022 London

Quantitative Analysis (2022)

Applied and advanced machine learning/AI and statistical learning methods for quantitative financial analysis
Utilized cloud and high-performance computing platforms to optimize data processing and analysis workflows

Awards

CDT Doctoral Scholarship in Biomedical Artificial Intelligence

UKRI 2020

Publications

Thomas Walker, Salvatore Esposito, Daniel Rebain, Amir Vaxman, Changjian Li, Arno Onken, Oisin Mac Aodha

January, 0001 In Arxiv

CrossSDF: 3D Reconstruction of Thin Structures From Cross-Sections

We propose CrossSDF, a novel approach for reconstructing a 3D signed-distance field from 2D cross-sections. The input is a set of 2D cross-sections that sample an unobserved ground-truth geometric object by planar intersection (denoted as black lines overlayed on the ground truth Alveolis structure on the left). CrossSDF (middle) accurately reconstructs thin structures without breakages, oversmoothing, or cross-sectional artifacts observed in competing methods (right).

Salvatore Esposito, Qingshan Xu, Kacper Kania, Charlie Hewitt, Octave Mariotti, Lohit Petikam, Julien Valentin, Arno Onken, Oisin Mac Aodha

January, 0001 In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) workshops

GeoGen: Geometry-Aware Generative Modeling via Signed Distance Functions

We introduce a new generative approach for synthesizing 3D geometry and images from single-view collections.This allows us to introduce useful priors to generate valid meshes. However those priors prevent the generative model from learning details limiting the applicability of the method to real-world scenarios. To alleviate that problem we make the transformation learnable and constrain the rendered depth map to be consistent with the zero-level set of the SDF. Through the lens of adversarial training we encourage the network to produce higher fidelity details on the output meshes.

Contact

Feel free to reach out

s2112162@ed.ac.uk
Informatics Forum, Edinburgh, Scotland EH8 9AB