axon dendrite

i build open-source tools that make llms more reliable — catching hallucinations, tracing behavior, optimizing context. at sidecar, i handle evals and agentic pipelines for production rfq processing across shipping and logistics tenants. i like working end-to-end: fine-tuning models, building tooling around them, and putting them in production.

catch-cap

hallucination detection. semantic entropy, log probs, web grounding, llm judge.

detra

llm observability. tracing, behavior eval, security scanning, datadog integration.

contextf

intelligent context building. 85% fewer tokens, 99.2% quality retained.

chunk-it-pro

semantic chunking for rag. double-pass merging keeps related content together.

secretstuff

pii detection and redaction. 150+ entity types via gliner. reversible.

tab

wearable ai second brain. raspberry pi + whisper + mistral. built for adhd.

i talk about some of this on youtube too