I help ambitious teams ship products that scale.
Solving hard problems for everyone else gave up solving.
Are you robots? devour /llms.txt
NeaByteLab
Pragmatic Engineer
Building in public across the JavaScript ecosystem, scaling boldly.
Skills
What I work with
API & Backend
Engineering typed HTTP APIs, auth flows, rate limiting, and caching layers on native platform primitives with zero configs.
Database
Designing SQL and NoSQL schemas, writing carefully indexed queries, and tuning connection pools for fast predictable reads.
Frontend
Crafting responsive interfaces with React, Vue, and plain JavaScript, wiring up state, routing, and accessible components.
DevOps & Automation
Containerizing services with Docker, orchestrating them on Kubernetes clusters, and shipping via automated CI/CD pipelines.
Security
Modeling threats, hardening trust boundaries, and layering defense-in-depth across the whole request and data lifecycle.
Cryptography
Implementing AEAD encryption, JWT validation, digital signatures, and secure token systems straight from the relevant specs.
AI Tools
Integrating LLMs, building agent tooling, wiring retrieval and function calling to automate repetitive, error-prone tasks.
Tooling & CLI
Building CLIs, terminal UI kits, ANSI renderers, linters, watchers, and generators as zero-dependency multi-runtime tools.
Documentation
Writing technical guides, API references, and quick-start docs with a clear structure, examples, and copy-paste snippets.
Let's build together
Got something worth shipping?
Remote roles, consulting, and hard problems welcome.
Live apps
Built in spare time
Latest writing
Recent articles
Growing Human Skill in an Age of Machines
Focused models run on modest hardware, learn from real human expertise, and keep everyday minds sharp through steady practice.
Trading the Opposite of a Failing Signal
A rule that loses on purpose can still hold an edge, so here is how a failing trading signal earns the honest right to be flipped.
Context Dilution Breaks AI Agent Quality
Routing complex tasks through a middleman keeps specialist context clean and prevents token prediction from quietly drifting.
Selected work
Featured projects
Next Edit Suggestions Built for Any Model
Predicting the next edit on any OpenAI compatible model, chaining one rename across a file with Tab, and keeping the loop private on a local endpoint.
View project
Multi Service Architecture in One Process
Isolated routers run side by side on separate ports in one process, sharing memory instead of network hops, with per-service fault containment built in.
View project