Syncro

Overview

Syncro is the SaaS I'm building now — the bridge between Solana's raw RPC/gRPC firehose and the Web2 relational database a product team already runs. Developers write a small AssemblyScript transform in a browser Monaco editor; the platform compiles it to sandboxed WebAssembly, spins up a dedicated Rust worker on Kubernetes, and that worker executes the Wasm per transaction and streams structured rows directly into the user's own Postgres — with schema migrations, backpressure and credit metering all handled for them.

The problem

Solana emits more data than any other chain — a relentless firehose of transactions over RPC and gRPC. But that raw stream is useless to a product team: there's no easy way to land it, in a shape they control, inside the database their app already speaks to. Today teams have two bad options. They rent rigid, opinionated indexers that trap their data behind a proprietary schema and API, or they hand-roll fragile Rust pipelines — babysitting RPC backpressure, reinventing batching, migrations, sandboxing and billing every time. Syncro deletes that entire middle layer: you describe the shape you want, point it at your own database, and the data shows up.

What I built

• A browser-based Monaco editor where developers write a simple AssemblyScript transform mapping on-chain accounts to their own schema — compiled server-side to a sandboxed Wasm binary and cached in Redis for instant pod boots.
• Generic Rust indexer workers (wasmtime) that boot from an INDEXER_ID, load the user's Wasm, serialize each Solana transaction via Borsh into linear memory, and stream the structured output straight into the user's own Postgres (Bring Your Own Database).
• A visual schema builder that generates versioned SQL migrations and applies them safely on pod boot via an _indexer_migrations_history table — additive-only, so a schema change can never silently wipe a user's data.
• A Spinner microservice that drives the Kubernetes API over RabbitMQ to spawn, pause and kill a dedicated worker pod per indexer, with Redis distributed locks preventing duplicate deploys on a double-clicked button.
• Backpressure-aware batching: workers batch rows in memory and throttle the Solana read-stream the moment the user's DB slows down — preventing OOM — with bb8/deadpool pooling shielding user databases from connection storms.
• A prepaid credit system with a Redis atomic DECR per transaction (no Postgres lock on the hot path) and Paddle-powered auto-recharge — Merchant-of-Record billing that bakes in global tax compliance for an India-based entity.

Architecture

Where it stands

In active development as my current flagship — a founder-led, full-stack build spanning a Rust control plane, a WebAssembly data plane, Kubernetes orchestration and a Next.js product surface. The hard, defining parts — Wasm execution, BYODB migrations, backpressure and high-speed credit metering — are exactly what I'm building right now.