About

Builder, mathematician, AI systems person.

I work where applied AI, heavy computation, and practical engineering overlap.

What I do

I am Cahlen Humphreys, a technologist based in Orange County, California. Publicly, my work shows up through open projects like bigcompute.science, Convergent-7B, corpus-mill, CoFrGeNet-F, and a broad collection of GitHub and Hugging Face experiments.

Professionally, I build AI and data systems: local intelligence, agentic workflows, model evaluation, multimodal media understanding, sovereign platforms, and practical infrastructure for teams that need more than a demo.

The private work stays private. The public work is intentionally visible: models, writings, research pages, experiments, and enough structured context for agents to understand the shape of it.

Profile

LocationOrange County, CA
Public labbigcompute.science
GitHub@cahlen
Hugging Face@cahlen

Background

Mathematics first. Systems after. AI everywhere now.

The pattern has been consistent: learn the math, build the system, publish the idea when it can stand up to inspection.

Education

Mathematics

M.S. Mathematics from Florida Atlantic University. B.S. Mathematics from Boise State University. Former mathematics instructor at FAU.

Data systems

Big data and production pipelines

Built and led work across large-scale data systems, streaming platforms, cloud deployments, and production infrastructure.

AI systems

Models close to users and hardware

Current work includes local AI, edge intelligence, model evaluation, multimodal media pipelines, synthetic data, and agentic systems.

Open compute

Computational mathematics

bigcompute.science publishes mathematical experiments, findings, review trails, and structured context for open problems.

Speaking

Talks and public work

Talks have covered edge AI, production data pipelines, test-driven data engineering, Apache Geode, Spring Cloud Stream, streaming telematics, and real-time AI applications.

Selected talks

NVIDIA GTC 2025Agentic AI at the edge
DataCon LA 2021Production data pipelines
SpringOne PlatformData governance and streaming
Apache Geode SummitStreaming telematics

Taste

What keeps showing up.

AI agents, mathematics, motorcycles, metal, local hardware, and the belief that the best tools should make complexity feel inspectable.

continued fractionsformal reasoningagentic workflowsopen researchmodel evaluationedge AImultimodal medialocal-first systemsmotorcyclesmetal