# Brandon Abreu Smith

> Co-founder @ Structured AI (YC F25) | Physics @ Oxford

Location: San Francisco, California, United States
Profile: https://flows.cv/brandonabreusmith

Currently building complex agentic systems at Structured AI. We are making use of frontier models and training new ones. Message me if you want to join us and work on the cutting edge. 

Previously:
- First authored NeurIPS paper with 30 citations on bias in VL models.
- Built first chunking evaluation framework for Chroma DB, reviewed by director of Stanford AI labs and gained 500 starts on Github.
- Founded 2 YC backed startups.

## Work Experience
### Co-Founder @ Structured AI
Jan 2025 – Present | San Francisco Bay Area
Making the AI workforce for construction design engineering.

### Software Engineer & AI Researcher @ Syska Hennessy Group
Jan 2024 – Present | London Area, United Kingdom
• Currently contracting for Syska Hennessy Group.
• Developed highly scalable chat bot, servicing hundreds of internal employees. Built on FastAPI, deployed in a docker container to a K8s cluster running on Azure. 
• Full DevOps setup for rapid CI/CD operations. Using Azure Pipelines to automatically update pods on K8s cluster based off pushes to main on the repo.

### Software Engineer @ Chroma
Jan 2024 – Jan 2024 | San Francisco, California, United States
• Wrote research paper on evaluating chunking algorithms for RAG. Reviewed by Chris Manning, director of Stanford AI Lab. 
• Project received over 50k views on X.com.
• Developed ClusterSemanticChunker and LLMSemanticChunker which achieved the highest recall across our evaluation set.

### Contract Software Engineer @ Syska Hennessy Group
Jan 2024 – Jan 2024
• Developing products for Syska's Innovations Program.

### Chief Technology Officer @ Project Chiron
Jan 2023 – Jan 2024 | San Francisco Bay Area
• Co-founded Project Chiron Inc. (C Corp, Delaware) as CTO; worked in San Francisco.
• Secured $47,000 from SF investors; discussed pre-seed with a16z. Networked with numerous investors.
• Developed Flask Python backend with Docker, hosted on GCP; auto-scaling with Cloud Run containers.
• Implemented DevOps: rapid development cycle with Cloud Build, GitHub triggers for automatic builds.
• Managed SQL server; optimized for 150,000+ rows of user data for efficient usage statistics algorithms.
• Implemented Redis cache for faster data delivery. Built Svelte frontend; integrated OpenAI API. Achieved 100 active users with average weekly usage of 80 hours. Implemented Stripe payment system.

### Software Engineering Consultant @ Syska Hennessy Group
Jan 2023 – Jan 2023
• Developed .NET Excel plugin and Azure Function to write weather data to a sheet.
• Using LangChain, developed Streamlit web app that uses LLMs to validate pdf submittals against a
criterion.
• Created React frontend and Flask backend for web app that forecasts cooling expenses for data centers.

### Machine Learning Researcher @ Oxford Artificial Intelligence Society
Jan 2022 – Jan 2023 | Oxford, England, United Kingdom
• First authored paper on assessing bias in Vision Language models, specifically CLIP.
• Programmed K-NN algorithm to filter poor quality images from augmented dataset built from generative AI. 
• Implemented various bias metrics and tested them on models.

### Software Engineering Consultant @ Joseph D'Angelo and Associates, Inc. (DNA)
Jan 2020 – Jan 2021
• Programmed inventory management software on Unity3D using C# that connects with Amazon Web Service (AWS) SQL databases. It exported Excel financial documents for use in projects with Facebook.

### Computer Lab Assistant @ Oxford Artificial Intelligence Society
Jan 2019 – Jan 2020 | Oxford, England, United Kingdom
• Assisted team in coding VR-experienced AIs.
• Developed an archery AI using TensorFlow and Deep Reinforcement Learning (DRL).
• Created an initial simple hide-and-seek AI, which was later expanded by the team.
• Utilized Unity3D with C# for AI simulation.
• Implemented the Proximal Policy Optimisation algorithm in DRL models.

### Software Engineer @ M2X.ai
Jan 2019 – Jan 2020
• Created TensorFlow Deep Reinforcement Learning (DRL) agent for optimized pipe placement, enhancing hallway space. Project achieved 1,000+ LinkedIn likes and attracted new clients.
• Conceptualized and coded a program in Python using OpenCV for counting symbols in architectural
blueprints. Further optimized it in C# for commercial release.


## Education
### F25
Y Combinator

### Master's degree in Physics
University of Oxford

### A-Levels in Maths, Physics and Further Maths
Sir William Borlase's Grammar School

### GCSE
Amersham School

### Z Fellows


## Contact & Social
- LinkedIn: https://linkedin.com/in/brandon-abreu-smith
- Portfolio: https://www.rededge.tech/

---
Source: https://flows.cv/brandonabreusmith
JSON Resume: https://flows.cv/brandonabreusmith/resume.json
Last updated: 2026-04-11