Researching and building practical, open-source AI systems.
🌍 Open to remote opportunities in software engineering, AI systems, and developer tooling.
TrainLens: ML training diagnostics with actionable next steps
TrainLens turns notebook training state into interpretable diagnostics, performance visualizations, and proposals for subsequent experiments. Its deterministic local analysis grounds recommendations in observed evidence. Optional LLM reports or coding-agent integrations extend the workflow.
ASTScribe: trace model setup, training, evaluation, and cell dependencies in ML notebooks
ASTScribe transforms supported Python ML code or Jupyter notebooks into evidence-backed methodological explanations. Static analysis identifies training or inference patterns, reconstructs cell dependencies, then traces findings to their original source lines—without executing code or relying on an LLM.
LastLight: offline knowledge retrieval with verifiable sources
LastLight makes locally stored knowledge accessible when connectivity or infrastructure is unreliable. Its standard-library Python engine searches Markdown or ZIP collections, exposes the evidence behind each result, then withholds answers when support is insufficient. No cloud services, embeddings or vector database required.
Semauri: deterministic programming through controlled natural language
Semauri explores a human-readable programming language with predictable compiler semantics. It generates web documents or JSON Schema, plans filesystem operations, and represents machine-learning workflows through typed constructs. Effectful actions require explicit authorization; compilation itself executes no side effects.
LIRUVA: Make music. Freely. — an open-source DAW built in Rust
LIRUVA is an early-stage, free and open-source digital audio workstation developed in Rust. Its development branches introduce a native desktop interface, a 16-step drum sequencer with real synthesized audio, and local project saving. Built for approachable, offline-first music production without subscriptions; it is still a prototype, not a full DAW.
Python Package Portfolio
Small, practical Python tools for AI evaluation, data quality, retrieval, multimodal analysis, visualization, and generative-audio workflows.
| Package | Research / engineering focus | Downloads |
|---|---|---|
| embedding-drift-lite | Embedding-distribution drift detection and diagnostic inspection. | |
| parametricbench | Provider-independent regression benchmarking for LLM and VLM systems. | |
| promptshield-llm | Prompt-injection, unsafe-instruction, and input-risk screening for LLM pipelines. |
| Package | Research / engineering focus | Downloads |
|---|---|---|
| metaclean-vlm | Metadata normalization for image and vision-language model datasets. | |
| visual-patch-audit | Image-patch auditing and visual-signal inspection for VLM workflows. | |
| vlm-occlusion | Grid-based black-box occlusion analysis for VLM claim sensitivity. | |
| vlm-prior-probe | Counterfactual black-box evaluation of whether VLMs follow visual evidence or learned priors. |
| Package | Research / engineering focus | Downloads |
|---|---|---|
| matplotlib-dark | Automatic dark mode for Matplotlib with ready-to-use themes and safe temporary styling. |
| Package | Research / engineering focus | Downloads |
|---|---|---|
| egypttranslit | Egyptological transliteration to clean Unicode from Python or the command line, with MdC conversion, normalization, validation, and diagnostics. |
| Package | Research / engineering focus | Downloads |
|---|---|---|
| text-to-music-prompt-structurer | Structured prompt extraction for text-to-music generation workflows. |
API Clients
Explore Cablegram, token-efficient communication for humans and AI agents
A research-oriented toolkit for minimizing token cost while preserving task-critical meaning through deterministic invariant checks, receiver-aware context, auditable candidate selection, and benchmarking.
Maximum meaning. Minimum tokens.
Explore Suffice, token-efficiency research for AI agents
Suffice measures how few tokens an AI agent needs to successfully complete a task across prompts, context, tools, memory, and final responses.
It searches for minimum successful token budgets and token-efficiency frontiers under explicit quality constraints, so reducing tokens only counts when task success is preserved.
Check MAVERICK, cognition-inspired AI research for multi-agent visual reasoning
Inspired by human visual cognition, MAVERICK decomposes image understanding into an auditable four-agent loop: perceive, describe, critique, and refine.
Designed for interpretable VLM workflows, uncertainty-aware reasoning, and stronger model-ready image descriptions.
Explore Neural Audio Theory, the engineering foundations behind AI music generation
An open educational project explaining how modern AI music systems work, from signal processing and embeddings to transformer and diffusion architectures, training, and prompt conditioning.
Built for developers and researchers who want a technical understanding of the systems behind neural audio and music generation.
Discover music-to-text, local-first AI tooling for turning audio into structured metadata and music industry copy
An open-source Python framework that extracts acoustic features and generates A&R notes, PR pitches, playlist descriptions, and sync licensing blurbs, with or without an LLM.
Built for reproducible audio analysis, structured exports, and OpenAI-compatible or fully local workflows.
Check my UI component library for AI
edujbarrios-ui is a frontend component library for building clean AI interfaces, demos, and developer tools.
Languages
AI Systems & Tooling
- I Created a Complete Educational Guide on AI Music Generation
- I Built My Own Documentation Site Builder - Here's Why and How
- I Built a Notebook Engine for C, and the Potential Is Incredible
OS & Development
