2026 to present
Independent
AI researcher & angel investor
Evolutionary systems research, open-source infrastructure, and early-stage AI investments.
RachPradhan
Researcher and open-source systems builder
- AI systems
- Developer infrastructure
- Singapore / San Francisco
I build tools for coding agents and study how software improves through search, feedback, and evolution.
My recent work spans code intelligence, package management, web frameworks, and AI infrastructure. I often use Zig at the systems layer, alongside Python and TypeScript. My open-source packages have been downloaded more than 1.2 million times.
Selected work
Six recent projects, with source and implementation details.
Agentic coding harness
codegraff
A coding harness in Zig with multi-provider support, MCP workflows, and a Darwin Gödel Machine evolution loop.
Evolution loopDarwin Gödel Machine
Code intelligence for agents
codedb
Indexes symbols, call sites, and dependencies so agents can inspect repository structure before reading whole files.
IndexesSymbols · calls · dependencies
Package manager for macOS and Linux
nanobrew
A native Zig package manager with parallel downloads, a static binary, and warm installs benchmarked against Homebrew.
Warm install230× faster than Homebrew
Web framework in Zig
merjs
File-based routing, SSR, type-safe APIs, and a WASM client without a Node runtime.
Client runtime0 Node.js required
Agent infrastructure
devswarm
An MCP server, code graph engine, and evolutionary algorithm platform with 33 focused tools.
Tool surface33 focused tools
FastAPI-compatible, Zig-powered
turboAPI
A drop-in FastAPI-compatible framework with a Zig HTTP core, native validation, zero-copy responses, and free-threading.
Benchmark20× faster than FastAPI
Project archive
All repositories ↗emergentDB
Vector database experiments with MAP-Elites
dhiSIMD validation for Python and TypeScript
BhumiAI inference client with provider adapters and structured outputs
SatyaSchema validation library for Python
kewA modern async task queue for Python
huewheelColor palette generator and search experiment
VroomSGBus tracking for 60k daily riders
condensation.aiEmbeddings with diffusion models
Research ideas, built into working systems.
I like research questions that survive contact with production.
That has taken me from autonomous mapping and medical imaging to AI infrastructure and tools for coding agents. I often work in Zig when control over memory and runtime behavior matters, but the language follows the problem.
I also make early-stage investments in AI infrastructure and developer tools, write about the experiments behind the work, and speak about evolutionary systems.
Talks
Research ideas, worked through in public.
Evolutionary Algorithms in Agent Behavior and Tooling
Agent traces, tool use, costs, errors, and outcomes become signals for search. The surrounding toolchain becomes the environment being optimized.
Experience
2025
Menlo.ai
Member of Technical Staff
Robotics systems and machine-learning infrastructure.
2023 to 2024
Condensation.ai
Founder
Diffusion models for disease detection in medical imaging.
2023
BGARAGE
Robotics intern
Developer tools and topometric mapping for autonomous systems.
Elsewhere
Cafe field notes
A map of cafes I return to in Singapore and San Francisco.
Visit ↗Technical writing
Essays on coding agents, model infrastructure, Zig, and open-source work.
Visit ↗One-North food guide
Twenty-one places for breakfast, lunch, dinner, and coffee around one-north.
Visit ↗Package name checker
Check PyPI and NPM availability, including batch lookups and name alternatives.
Visit ↗Contact
If you are building coding agents, developer infrastructure, or research that needs to ship, I would like to hear from you.