Projects

Posted on Jun 10, 2023

Here are a few projects and experiments I’ve worked on. I’ll keep expanding this list over time.

Redistill case study

An extremely fast Redis-compatible key-value database focused on low-latency production workloads. In the published single-instance benchmark, it reached 9.07M operations per second with 0.479 ms p50 latency.

My role: I designed and built the entire system, including the database engine, Redis protocol support, persistence, operational controls, packaging, documentation, and performance work.

Why it matters: Redis-compatible systems are often used in the hot path. Redistill demonstrates what a focused, multi-core architecture can achieve while preserving compatibility with existing Redis clients.

Area: Databases, distributed systems, performance engineering

Read the case study · Visit the website · View the source

AM2 case study

A macOS terminal dashboard that brings local coding agents into one view. It discovers Codex, Claude, OpenCode, Pi, Gemini CLI, and Cursor Agent processes, with optional Herdr integration for pane status and navigation.

My role: I built the Rust terminal application, including agent discovery, project views, session history and search, terminal navigation, and agent launching.

Why it matters: Running several agents across terminals makes it easy to lose track of where each one is working. AM2 gives those sessions a shared, local view without requiring Herdr for discovery.

Area: Developer tools, AI agents, terminal applications

Read the case study · View the source · Download for macOS

MemX case study

A local-first AI memory layer for engineers working across tools like Claude, Cursor, ChatGPT, and other agents.

My role: I designed and built the entire product, including its storage model, local AI pipeline, semantic retrieval, CLI, MCP server, and agent integrations.

Why it matters: AI tools lose context easily. MemX keeps summaries and atomic facts in a local SQLite store, then makes them available across agents through MCP.

Area: AI agents, local-first tools, developer experience

Read the case study · Read the technical article · View the source

Atlas

An open-source AI answer engine.

Why it matters: Answer engines are a practical way to explore retrieval, reasoning, and user-facing AI workflows without hiding the system behind a black box.

Area: Applied AI, search, answer generation

Mumbai AQI

A community-driven air-quality tracker and recommendation platform.

Why it matters: Useful civic software should make local data easier to understand and act on, especially when the data affects everyday decisions.

Area: Data platforms, public-interest software, recommendations

More experiments