Tools I ship with
Here's what I reach for, and why.
AI / Vision
Computer vision, machine learning, and AI integration, from real-time image processing to production inference pipelines.
- PyTorch
- My primary deep learning framework. Used for object segmentation, pose prediction, and training custom vision models in production.
- TensorFlow
- Deep learning for image classification, object detection, and deploying models via TensorFlow Serving and TFLite.
- OpenCV
- Image processing, feature detection, and real-time video analysis. The foundation of my custom vision pipelines.
- Hugging Face
- Pre-trained transformer models for NLP, image segmentation, and domain-specific fine-tuning on custom datasets.
- ONNX
- Cross-platform model deployment: train in PyTorch, run optimized inference anywhere.
- OpenAI
- GPT and multimodal APIs for natural language features, document understanding, and automation.
- scikit-learn
- Classical ML for tabular data, feature engineering, and lightweight predictive models that don't need a GPU.
- MLOps
- Model lifecycle management: automated training, artifact collection, versioning, and deployment monitoring.
Frontend
Web interfaces built for speed, accessibility, and real-time data, from static marketing sites to complex annotation platforms.
- React
- Component-driven UI for interactive dashboards, annotation tools, and complex client applications.
- TypeScript
- Type safety across the stack, for fewer runtime bugs and easier refactoring.
- Next.js
- Full-stack React framework for SSR, API routes, and applications that need SEO and performance.
- Vue.js
- Lightweight reactive framework, good for embedding interactive features into existing applications.
- Astro
- Zero-JS-by-default static sites with island architecture. What this site is built on.
- Tailwind
- Utility-first CSS for fast, consistent styling without fighting specificity.
- WebRTC
- Real-time video streaming for live camera feeds, pose prediction, and peer-to-peer communication.
Backend
APIs, services, and server-side logic, including the services that run AI workloads.
- Python
- The language most of my AI/ML work is written in: data pipelines, model serving, image processing, and automation.
- Node.js
- Event-driven runtime for APIs, real-time services, and serverless functions, with a package for nearly everything.
- C#/.NET
- Backend for high-throughput APIs, WPF desktop apps, and legacy system modernization.
- ASP.NET
- Web APIs and MVC applications for enterprise environments with role-based access control.
- FastAPI
- High-performance Python API framework with automatic OpenAPI docs. My default for serving ML models.
- Express
- Minimal Node.js server for REST APIs, webhooks, and lightweight microservices.
Data
Persistent storage, caching, and analytics, with the database chosen to fit the workload.
- PostgreSQL
- My default relational database. JSONB, full-text search, and decades of reliability.
- MySQL
- Relational database I've used in production ML platforms and web applications.
- SQL Server
- Enterprise data platform for .NET ecosystems, reporting, and legacy system integration.
- Redis
- In-memory cache and message broker for session storage, rate limiting, and real-time leaderboards.
- MongoDB
- Document store for unstructured data, rapid prototyping, and content-heavy applications.
- ETL Pipelines
- Custom data ingestion and transformation workflows: cleaning, normalizing, and loading data for analytics and ML training.
Tools & Infra
Deployment, CI/CD, and development infrastructure, automated and reproducible from day one.
- Docker
- Containerized builds for consistent dev/prod environments and easy deployment of ML models.
- AWS
- EC2, Lambda, S3, and SageMaker for compute, storage, and model hosting.
- Cloudflare
- Edge hosting, Workers, Pages, and DNS. This site runs on it.
- Vercel
- Zero-config frontend deployments with edge functions, preview branches, and instant rollbacks.
- GitHub Actions
- CI/CD pipelines for automated testing, linting, building, and deployment on every push.
- Linux
- Production server environment. Comfortable with Ubuntu, Debian, Alpine, and custom AMIs.
- Jupyter
- Interactive notebooks for data exploration, model experimentation, and client-facing analysis reports.
- Git
- Version control with clean branching strategies, code review workflows, and automated releases.
Bring me a hard problem.
Project-based engagements and part-time retainers. No full-time contracts.