I build AI systems, agent workflows, and developer infra.
Lately I’ve been deep in Claude Code, Pi Agent, codex, subagents, retrieval-first context, and turning messy workflows into fast, LLM-native tools.
Been building since summer of 2016.
- Languages: Python, Go, Bash, JavaScript, Rust
- Infra & Ops: Kubernetes, Terraform, Helm, Docker, Ansible, Flux
- Cloud: AWS, GCP, Azure
- Data & Observability: PostgreSQL, MySQL, Redis, MongoDB, Prometheus, Grafana, Loki, Tempo, VictoriaMetrics
- Pi Agent, Claude Code, codex, OpenRouter, Hermes Agent
- Subagents, clean and precise context every time, SKILL.md > MCPs for most cases.
| Project | Description |
|---|---|
| xray | structural code intelligence for AI assistants using ast-grep. |
| chorus-ai | Frontend-only LLM ensemble interface with bring-your-own OpenRouter/API keys and client-side multi-model workflows. |
| chat-harvest | Local-first archive for AI chat traces across agents, designed as a substrate for DSPy/GEPA-style optimization loops. |
| ita-kg | Knowledge-graph representation of the Indian Income Tax Act so agents can traverse section-to-section relationships explicitly. |
| hashlines-mcp | MCP server for hashline-anchored safe file read/edit workflows inspired by this archived write-up: https://web.archive.org/web/20260412162933/https://blog.can.ac/2026/02/12/the-harness-problem/ |
| contextgraph | Decision audit ledger for AI agents - captures "why" as queryable data. |
| alpha | IAM policy rightsizing agent with AI-powered risk signals and instant rollback. |
| Project | Description |
|---|---|
| kubectl-smart | turns noisy Kubernetes debugging into signal-first workflows. |
| logsieve | log deduplication sidecar using Drain3 algorithm, reduces volumes ~90%. |
| scx-slo | eBPF CPU scheduler enforcing latency SLOs via Earliest Deadline First. |
| softcron | Kubernetes CronJob controller with Prometheus gating, pacing, and jitter. |
- Slow down Disk I/O → Flash talk on preventing
rmfrom nuking SSDs, deep dive intoionice,rsync, and cgroups v2.
💬 Reach me on LinkedIn or check out my projects here.
| Project | Description |
|---|---|
| podcast-watcher | Open-source attempt to replicate Gemini-style long-form YouTube understanding: extract MP3 audio with ffmpeg, build timestamped transcripts, detect only meaningful frame changes, and feed the useful video/audio context into an LLM. I found the visual path added little value, so it effectively converged into an audio-first podcast/video understanding pipeline. |
| netwatchrs | Rust network diagnostics toolkit I built while debugging inconsistent Wi‑Fi and suspected service-quality issues on my own machine; the goal was to see what was actually happening under the hood instead of relying on vague network status indicators. |
| murmur | Wispr Flow OSS - live streaming voice-to-text using whisper.cpp with Metal acceleration. |
| wiki-in-a-box | offline Wikipedia with hybrid no‑index RAG - powered by gpt-oss:20b |
| claude-memory-viz | memory visualization for Anthropic's Claude memory MCP. |
| vigil | Watch your codebase breathe; TUI for real-time file change monitoring with inline diffs |
| extract-lessons-book | Ebook/document normalizer with optional OCR and LLM-generated advice + flowcharts. It also became a practical OCR evaluation project for me: for most books, plain PDF text extraction worked better than heavy OCR, so GLM OCR ended up as a fallback reserved for noisy pages, scans, or extraction failures. |