My exact toolkit
hardware, software & workflow

Daily driver for all heavy development, ML training, Docker stacks, and pipeline work. Native Linux means zero WSL overhead — everything runs at full speed.

Used for client calls, presentations, and on-the-go work. Seamless handoff between machines via cloud sync.

Running multiple Docker Compose stacks, Jupyter Lab, VS Code with heavy extensions, and browser sessions simultaneously without swapping.

Fast I/O matters when working with large datasets, model checkpoints, and container images.

High DPI for long coding sessions. Side-by-side terminal + browser without strain.

Oh My Zsh with custom prompt on Ubuntu. iTerm2 on macOS. Multiple panes for dev server, file watcher, and logs running in parallel.

With ESLint, Prettier, Tailwind CSS IntelliSense, GitLens, and GitHub Copilot. Theme: One Dark Pro.

All exploratory data analysis, model prototyping, and notebook-based research runs here.

Every project is version-controlled from day one. GitHub Actions for CI/CD pipelines.

All production ML stacks (FastAPI, MLflow, Prefect, Grafana) deployed as multi-service Compose stacks.

Testing REST APIs during development. Collections shared across projects.

Linux toolchain on Windows. All bash scripts, cron jobs, and shell-based pipelines run here.

Oh My Zsh with custom prompt. Multiple panes for running dev server, watcher, and logs simultaneously.

All data pipelines, model training, FastAPI backends, and ETL scripts.

PyTorch for research and fine-tuning. Keras for rapid prototyping and deployment-friendly models.

Production workhorse for tabular data. XGBoost with Optuna Bayesian hyperparameter optimization.

RAG pipelines, GraphRAG knowledge graphs, and multi-agent workflow orchestration for enterprise clients.

GPT-4o for reasoning tasks. Embedding models for vector search in RAG systems.

Every training run logged with hyperparameters, metrics, and model artifacts. Model registry for versioning.

Scheduled data ingestion, automated retraining, and batch prediction workflows.

Waterfall and beeswarm plots for every production classifier. Explainability is non-negotiable.

Production database for all full-stack applications. JSONB for semi-structured data.

Prototyping and local pipeline development. Portable and zero-config.

Pandas for most work. Polars when performance on large frames matters.

All intermediate pipeline artifacts stored in Parquet. Columnar, compressed, fast.

ML model drift monitoring, API latency tracking, and data quality dashboards.

This portfolio. Server Components, SSG, ISR, and API routes — all in one framework.

Utility-first CSS. No more context-switching between files for simple style changes.

All scroll-reveal and micro-interaction animations. Always wrapped with useReducedMotion() for accessibility.

Unstyled, accessible components. Copy-paste into your codebase and own the code.

Strict mode enabled on every project. Catches bugs at compile time, not in production.

Primary cloud for client deployments. S3 for model artifact storage, EC2 for compute-heavy workloads.

Zero-config Next.js deployment. This portfolio lives here. Edge Network + automatic CI/CD on push.

All server administration, cron jobs, and shell scripting. Red Hat Certified Engineer.

Enterprise network design. 11 years of hands-on infrastructure from healthcare to government scale.

Contact form email delivery. Clean API, great deliverability, Next.js native integration.

Used daily for complex reasoning tasks, code review, architectural discussions, and writing. Significantly better than alternatives for systems-level thinking.

Supplementary to Claude for specific tasks. OpenAI API used in production RAG pipelines.

All project documentation, client meeting notes, and research notes live here.

Wireframing and design review for client-facing applications.

All 13 global client consultations. Zoom for US/EU clients, Meet for others.

Issue tracking and sprint planning for client projects.