Software Engineer @ Microsoft Azure 2023 – present
GPU cluster management (500k+ NVIDIA H200/GB200). LLM inference engines for Copilot. ML platform infrastructure at petabyte scale.
Diego Iruretagoyena
Software Engineer · Microsoft Azure High Performance Computing · diegoiru@stanford.edu / diiruo@gmail.com
I care deeply about creating products that make people's lives better. I'm passionate about machine learning, engineering, and societal wellbeing.
Experience
Founder in Residence @ Entrepreneur First 2022
AI chip design (Astrus) and healthtech marketplace (Gotta). 50+ customer interviews, 3 government grants.
Software Engineer @ ZeroFox 2021
Cybersecurity cloud infra. Distributed Go microservices, Terraform, AWS cost optimization (−75%).
AI Engineer @ Twenty Billion Neurons 2020
Deep learning video processing for human movement recognition. Company advised by Yoshua Bengio, acquired by Qualcomm.
Software Engineer Intern @ Amazon 2019
Supply chain optimization algorithms for the EU logistics network. Zero-carbon routing engines.
Software Engineer @ Fintual YC S18 2018
ETF visualization engine (D3.js, Python). Interactive financial dashboards for client conversion.
Education
Stanford University In Progress
AI Graduate Certificate (SCPD). CS231N Deep Learning for Computer Vision, CS229 Machine Learning, CS230 Deep Learning. Corporate sponsorship by Microsoft.
ML Research Scholar. Deep learning on ConvGANs with Alex Dimakis. First undergrad awarded the AI Research Scholarship.
B.Sc. Computer Science, Minor in Statistics. Chapter author, 'ChatGPT und andere Quatschmaschinen' (2024).
Projects
EEG-Based Recommender System
Pontifical University of Chile · Recommender Systems Lab
Built a recommender system driven by electroencephalography (EEG) brain signals. Used the DEAP dataset (32 subjects, 32 EEG channels, 128 Hz) to generate topographic brain maps via the MNE library. Extracted neural features from subjects watching video stimuli, constructed user-item interaction matrices, and applied collaborative filtering (implicit library) to predict content preferences from brain activity. Evaluated with MAP, nDCG, and precision@k.
![[Poster]](https://github.com/Iruretagoyena/RecommenderSystemEEGSignals/blob/main/Poster.png){kind=link}
LLM Inference at Scale
Microsoft Copilot · Azure High Performance Computing
Designed and operated an aggregate-merge inference engine serving millions of prompts per hour across 500k+ NVIDIA H200/GB200 GPUs. Built distributed systems for model sharding, request routing, and load balancing with sub-millisecond tail latency. Implemented GDPR-compliant data handling pipelines including prompt redaction, geo-fenced inference routing, and audit logging across EU and US regions. End-to-end telemetry with Prometheus, Grafana, and custom anomaly detection for SLA enforcement.
Ocean Drifter — Computational Fluid Dynamics & 3D Printing
Pontifical University of Chile · Fluid Mechanics
Designed and 3D-printed a Lagrangian ocean drifter for measuring surface currents. Modeled hydrodynamic drag and vortex shedding using computational fluid dynamics simulations. Fabricated custom components (blades, cylinder, housing) via FDM 3D printing, iterated on geometries to minimize drift error. Validated against in-situ oceanographic measurements.
Cybersecurity Incident Map Visualization
ZeroFox · Data Visualization
Built an interactive choropleth map for visualizing cybersecurity incidents at the district level. AngularJS frontend with D3.js and TopoJSON for geographic rendering, aggregating real-time threat intelligence feeds into spatial heatmaps. Enabled analysts to identify regional attack patterns and allocate response resources.
Deep Learning for Human Movement Recognition
Twenty Billion Neurons · Company advised by Yoshua Bengio
Developed real-time video recognition pipelines for an AI fitness coach. Implemented temporal convolutional networks (TCN) and 3D-CNN architectures in PyTorch for classifying human movements from video streams. Built preprocessing pipelines for optical flow extraction, pose estimation, and temporal segmentation. Optimized inference for edge deployment with model quantization and TensorRT. Company acquired by Qualcomm.
Founding & Ventures
Astrus — Physics-aware AI chip design EF, 2022
Gotta — Blood donation marketplace EF, 2022
Eredità — Sports league platform, $100k ARR, 800+ weekly users 7 years
Interests
Offline, I enjoy backpacking, free diving, and road cycling. I've been fortunate to be able to experience the world through my work. I've visited 40+ countries, made countless friends along the way. I free dove with humpback whales and toured Shenzhen and Beijing for free thanks to a cold-email. Some of the photos of these experiences can be found here.