I build production Solana smart contracts (Rust / Anchor 0.30+), high-throughput LLM gateways, and fail-closed AI agent spend guardrails backed by automated tests and 22 merged upstream open-source PRs.
Scoped milestone packages or $60/hr execution via Upwork or direct contract.
Custom Token-2022 transfer hooks, PDA state vaults, CPI guards, on-chain DEX log parsers, and typed TypeScript client SDKs with localnet tests.
On-device-first hybrid routers, semantic/exact caching proxies, multi-provider failover, and atomic per-tenant budget kill switches.
Resilient BullMQ + Redis task pipelines, fail-closed agent checkout flows, raw-body webhook verifiers, and restart-safe state recovery.
Production-tested TypeScript repositories with commercial deployment licenses delivered via Polar + Stripe.
BullMQ + Redis LLM queue worker with Lua-backed budget kill switches, OpenAI/Anthropic failover routing, and 7/7 passing Vitest checks.
Complexity-classified routing blending on-device inference with cloud fallback, PII protection, and 11/11 passing Vitest checks.
Bounded AI purchase intents for PayPal Orders v2 with strict JSON schema validation, synchronous reservations, and raw-body RSA webhook verification.
github.com/sidsri14/paypal-scopeguard →Pump.fun instruction decoding, constant-product bonding curve simulation, Raydium CPMM swap math, and MCP server tools.
github.com/sidsri14/solana-dex-analytics →Forge MCP circuit breaker protecting multi-agent tool calls with bounded budgets, retries, and clean fail-closed enforcement.
github.com/sidsri14/atlassian-rovo-guard →Routing routine and PII prompts locally before escalating complex reasoning to cloud LLMs with bidirectional circuit breakers.
Read Article →How atomic token accounting and queue circuit breakers keep production AI workers bounded under burst retries.
Read Article →Why HTTP 200 status codes are insufficient when caching SSE byte streams in Rust reverse proxies.
Read Article →