SyberLabs
empirical & experiential systems
by mateo robles
the operators
Principles of Design
01

Stochastic

→ Unpredictability
noise · probability · surprise
Vital Language ↗
02

Dynamical

→ Points in Time
attractors · feedback · regime shifts
Dynamic Kernel ↗
03

Generative

→ Articulation
semantic fields · steering · decoder boundary
Dreamer v2 ↗
04

Geometric

→ Points in Space
symmetry · topology · proportion
Eidolon ↗
05

Experiential

→ Subjectivity
audio · perception · presence
RISE ↗
the practice

My principal interest lies in the question: how may software change what people can perceive, understand, and create? SyberLabs is my place for building experiments across machine-learning evaluation, simulation platforms, and generative art. I graduated from Santa Clara University in Computer Science, with an emphasis on Data Science.

Strength
Turning ambitious ideas into testable software systems.
Range
ML, simulation, generative art, research tooling.
Signal
Evidence-aware framing: results, prototypes, and art stay clearly labeled.
selected work
Built & Tested

SkyPredict

Applied ML

Flight-delay prediction built around strict pre-departure observability, staged feature pipelines, and an audited reporting dashboard.

  • 6.96M BTS flight rows under gate-close causal constraints.
  • Stage C validation: ROC-AUC 0.713, PR-AUC 0.349 at validation prevalence.
  • Work centered on pipeline architecture, leakage audits, dashboarding, and reporting.
PythonML pipelineEvaluationDashboard

Dynamic Kernel

Simulation

A simulation and API platform for non-stationary Markov processes over weighted graphs, with claims tied to versioned scripts and frozen artifacts.

  • Vectorized NumPy kernel, population simulator, FastAPI/WebSocket service, React/Vite visualizer.
  • 145-test public reproduction suite, plus falsification reports and archived pre-fix estimator artifacts.
  • Includes a semiconductor policy demo as an inspectable evidence slice.
PythonFastAPISimulationEvidence tooling

RISE

Experience

A self-contained audiovisual web application with its own audio engine and generative visuals — software built to be encountered, not just operated.

  • Built solo with custom rendering and audio systems.
  • Real-time audio and procedural visuals kept in sync in the browser.
  • The studio's most-developed front-end application.
Creative codingAudioCanvasWeb

Grokking Scaling Theory

Research

A pre-registered empirical ML study of delayed generalization, extended into categorical order parameters and the wave code result.

  • 63 deterministic modular-arithmetic runs across five architectures, three moduli, and trace-based diagnostics.
  • Tests sheaf/kNN geometry and logical-cell hypotheses; finds a leading sheaf signal while logical-cell assembly remains unsupported.
ML scienceGrokkingOrder parameters

Dreamer v2

Generative systems

A semantic-dynamics engine that inserts a testable intermediate representation between a prompt and generated language, with the decoder kept as an explicit boundary.

  • Pure engine path runs without API keys; optional prompted and activation-steering decoders.
  • Steering diagnostics found shared-direction collapse, then added decorrelation and query-dependent seeding.
GenerationSteeringDiagnostics

Vital Language

Research

A controlled generation study testing whether chaotic logit modulation could make LLM prose feel more alive. The result sharpened the hypothesis: coherence scaffolding mattered more than token turbulence.

  • Matched chaos against OU-noise, white-noise, sampling, and prompt-agency controls.
  • Negative result preserved: the wrong lever was part of the finding.
LLMControlsComplexity metrics

Eidolon

Digital art

Strange attractors, Bessel-mode resonances ("frozen music"), and tensegrity normal modes rendered as light: every visual derived from a verified invariant.

  • Browser-native works built from real mathematical systems.
  • Math verified before it is drawn.
CanvasGenerativeMath

Green Hypercube

Data science

A coverage-aware computational ethnobotany benchmark for testing structured search over plant data without mistaking database attention for biological signal.

  • Decoupled reward provenance, negative controls, and matched-density sweeps.
  • Findings separate raw cue-reward coupling from study-effort artifacts.
EthnobotanySampling biasReproducibility

Clarity Engine

Research tooling

A research engine that maps ambiguous problems into causal structures, tracks provenance, and keeps hypotheses, simulations, and evidence behind explicit review gates.

  • Tree of Knowledge graph for structural signatures, provenance, and causal-coordinate retrieval.
  • Research workflows keep candidates, simulations, and evidence transitions deliberately separated.
Knowledge graphCausalityEvidence gates
professional directions
Three ways I can contribute

Research engineer

Grokking Scaling · Dynamic Kernel

I build empirical machinery for hard-to-test ideas, and report uncertainty as part of the work.

ML / data engineer

SkyPredict · Green Hypercube

I build end-to-end data systems with observability, evaluation, audits, and interfaces.

Creative technologist

Eidolon · RISE · Clarity Engine

I turn mathematical systems into usable, interactive software with premium visual design.

Looking for teams building useful software with care.

contact github resume