AI engineering · open source · developer tools
I build and write about AI engineering, open-source infrastructure, and developer tools for people who like shipping. I like turning messy systems into tools, notes, and experiments other builders can reuse.
junaidshaukat.com · Blog · X · LinkedIn · Email
- Google Summer of Code 2026 with the Apache Software Foundation — building a portable Kafka Streams runner for Apache Beam. Translates Beam pipeline protos into Kafka Streams topologies against the Fn API, so pipelines written in any Beam SDK run on it.
- Kubernetes Release Team — Release Signal Shadow for the v1.37 cycle. Tracking CI signal, flaky tests, and release-blocking issues across SIGs.
- Writing about LLM tooling and agentic workflows.
| Project | What it is | Traction |
|---|---|---|
| isDisposable | Disposable-email detection API and offline package | 23k+ npm · 850+ PyPI · 160k domains indexed |
| reqcraft | 3 kB zero-dependency axios alternative on native fetch |
Built the week of the axios supply-chain attack |
| QueryMaster | MCP server for querying any SQL database in natural language | — |
| revise-deps | CLI to audit and update Node.js dependencies | Published on npm |
| Storage Sweep | Native macOS storage cleaner, SwiftUI | Frees 25–90 GB per machine |
| Project | Work |
|---|---|
| apache/beam | GSoC 2026 — portable Kafka Streams runner (PRs) |
| kubernetes/kubernetes | Release Signal Shadow, v1.37 (sig-release) |
| apache/airflow | PRs |
| archestra-ai/archestra | PRs |
| google-gemini/gemini-cli | PRs |
| webpack/webpack | PRs |
| jaegertracing/jaeger | PRs |
| InsForge/InsForge | Core plus JS, Swift, and Kotlin SDKs (PRs) |
| CCExtractor/ccsync | PRs |
| PalisadoesFoundation/talawa-admin | 6 merged PRs — i18n components, tree-shaking, 30% smaller bundle |
Plus mainline Linux kernel patches on ARM64, submitted the old way with
git format-patch and git send-email.
- How to Train a Small LLM Without Burning Money
- More Context Is Making Your AI Dumber
- Agentic Loop Engineering: The Missing Infra Layer
- I Built the Same AI Agent in Four Frameworks. Here's What I Learned.
- I Built a Zero-Dependency Alternative to axios After the Supply Chain Attack
AI engineering — LLM APIs (Anthropic, OpenAI, Gemini), MCP, tool calling, agent loops, LangGraph, RAG, embeddings, pgvector, evals and LLM-as-judge, LoRA / QLoRA fine-tuning, vLLM, Ollama
Languages — TypeScript, Python, Java, Go, C/C++, SQL, Bash
Backend & data — Node.js, Express, GraphQL, PostgreSQL, MongoDB, Redis, Kafka
Infrastructure — Docker, Kubernetes, AWS, GCP, Linux, Supabase
Frontend — React, Next.js, Tailwind
I work inside agentic dev tools daily — Claude Code, Codex, Cursor, Zed — and compose them into pipelines rather than using them one prompt at a time.
junaidshaukat.com · junaidshaukat546@gmail.com