Data Science | Applied AI/ML | Software Engineering

Marcos Vieira

I build reproducible data, applied AI/ML, and software systems that turn experiments into validated, documented, and decision-ready workflows.

View Case Studies Research Background
Software Full-stack MVPs, APIs, frontend interfaces, Docker, CI/CD, tests, documentation
Data Science And Applied AI/ML Classical ML, computer vision, NLP, LLM/RAG, metrics, and model evaluation
Research Engineering scRNA-seq, stochastic modeling, sensor data, scientific computing, statistical validation

Selected Evidence

Portfolio Work

The portfolio is organized around evidence: shipped software, reproducible data and ML workflows, peer-reviewed research engineering, scientific computing, and technical writing that explains decisions and limitations.

Software Engineering | Public MVP

The Tavern's Ledger

A full-stack cataloging system that demonstrates API design, a dynamic SPA frontend, Docker, tests, CI/CD, and design-first documentation.

  • RESTful Flask backend and vanilla JavaScript SPA frontend.
  • API configuration used as a single source of truth for frontend behavior.
  • Documentation, tests, Docker, and CI/CD treated as part of delivery.
BackendFrontend

Data Science / Applied AI/ML | Active track

Operational AI Field Notes

Reproducible applied AI experiments and technical write-ups focused on model behavior, evaluation, service boundaries, and operational contracts.

  • Classical ML, computer vision, NLP, and LLM/RAG experiments.
  • Notebook-to-service refactors with tests and explicit evaluation boundaries.
  • Focus areas include calibration, text intelligence, retrieval evaluation, and operational decision workflows.
Repository

Research Engineering | Peer-reviewed

GBM scRNA-seq Attractor Analysis

Open-source research engineering around single-cell RNA sequencing, dimensionality reduction, clustering validation, and stability-oriented gene regulatory network analysis.

  • First-author Scientific Reports paper and related IJMS publications.
  • High-dimensional biological data workflows with statistical validation.
  • Bridges computational biology, scientific software, and public research artifacts.
RepositoryScientific Reports

Scientific Computing | Published research

Timepix Pixel-Detector Pipelines

Sensor-data workflows for radiotherapy photon-beam spectrum reconstruction, including calibration, simulation, event reconstruction, and uncertainty-aware validation.

  • Pixel-level data processing and custom C++/ROOT algorithms.
  • Simulation, inverse-problem reasoning, curve fitting, and chi-square validation.
  • Strong foundation for data analysis, scientific computing, and uncertainty handling.
RepositoryPublication

Positioning

From Research To Delivery

My background connects applied physics, computational biology, scientific software, and practical data/AI workflows. The portfolio is designed for software, applied AI/ML, data science, analytics, and data engineering roles where evaluation, reproducibility, technical communication, and handoff-ready code matter.

Formless Solutions extends this work into an emerging consulting studio focused on applied AI, data and software systems, strategy, and decision workflows for complex problems.

Applied AI/ML Focus

Current Case Study Directions

The active AI/ML portfolio track focuses on practical workflows that make model behavior visible: calibration, text intelligence, retrieval quality, evidence grounding, and business-facing interpretation.

Computer Vision

Face Recognition Calibration

Thresholds, FAR/FRR, open-set limits, and why calibration is not liveness detection.

NLP And Analytics

Complaint Intelligence

Text signals, sentiment limits, aspect cues, and operational triage over public complaint data.

LLM/RAG

Contract RAG Evaluation

Citations, abstention, retrieval quality, and auditable document intelligence.