>_ diego_salazar_
Back to Home

:: MODULE: LAB

Technical Experiments

Where I synthesize theory into functional architecture. Applied research, open source benchmarks, and RAG pipelines.

Latest Build
v1.0.4 (Illari)
Core Stack
Next.jsPythonQdrant
System Online

Illari

RAG Pipeline & Safety Evaluation

v1.0.4
operational
Next.js
Python
Qdrant
LangChain

The Hypothesis

"Can we reduce hallucination rates in legal query retrieval by implementing a dual-stage verification layer based on epistemic uncertainty metrics?"

The Outcome

Achieved a 14% reduction in non-factual citations compared to baseline naive RAG implementation. Released as open-source benchmark.

Note

This safety protocol is currently deployed in production for select clients via my Advisory practice.

Andes Runners

High-altitude Performance Modeling

v2.1.0
operational
PyTorch
React
TimescaleDB

The Hypothesis

"Does incorporating geolocation altitude metadata into physiological time-series data improve the prediction accuracy of cardiovascular strain?"

The Outcome

Developed a predictive model used by 3 local athletic clubs to adjust training loads dynamically.