I'm Anirudh. I build at the intersection of AI and crypto — real-time AI tools that work in daily life, and onchain systems that hold up under production pressure. Based in Bangkok.
This is where I document what I'm building, why I built it, and what's shipping now.
AI
Basic prompts were producing vague, generic outputs. Built PromptBetter to sit in front of Claude Code and rewrite prompts for clarity and technical depth before anything runs. Using it daily and improving it — better rewrite quality, tighter approval flow.
I built Saphan because daily conversations in Thailand became hard when language got in the way. Most translation tools I tried felt robotic in real human conversation. I am improving real-time voice quality, reducing latency, and making turn-taking feel more natural during long conversations.
Grew up in the gems and jewelry industry. Traders run entire businesses on Excel and WhatsApp — no real data, no visibility, no way to make informed decisions. Built this because I understood the problem from the inside. POC is done. Starting to demo to first clients soon. Built everything solo — product, requirements, and code.
Crypto
First real exposure to production DeFi. Built automation bots for strategy rebalancing and smart contract operations — learned fast what happens when onchain automation breaks in a live system. Also built the ops dashboard for monitoring contract state, tracking transactions, and managing protocol deployments. Got a practical education in how yield strategies actually execute across chains, and how to instrument systems so you know when something is wrong before users do.
Started as an intern, ended up doing backend infrastructure and developer relations across two and a half years. On the infra side: built TRON gasless transaction infrastructure using JIT energy rental that cut costs by 60%, and designed the status API that handled 3M+ transactions a month across 50+ chains. On the devrel side: answered thousands of developer questions from 30+ protocols — MoonPay, Kraken, Infinex, MakerDAO, Beefy, and others — and wrote docs from scratch when official ones didn't exist. Also helped enterprise clients like Kraken, Lisk, and Reya figure out their chain architecture: DA layer choices, sequencing, settlement tradeoffs. That work taught me something most backend roles don't: what it means to build for developers, not just users. You understand failure modes differently when you're also the one explaining them.
Also built an elizaOS AI agent during that time — presented it live at ETH Dubai 2025. open source · watch the demo.
Built to explore x402 micropayments and verifiable randomness. Wanted to see if per-request payment for randomness made sense for autonomous agents — no subscriptions, no pre-funding, no API keys. Experiment is done. Not actively developing.
Now
- Demoing the jewelry trade ERP to first clients
- Iterating on PromptBetter based on daily use
- Publishing build logs as things ship
Writing
March 2, 2026
before you build with ai, you should probably understand what it's doing
January 10, 2026
why i built a real-time translator
July 29, 2025
Fixing Size Recommendations for Online Clothing
December 8, 2024
Simple Steps
October 20, 2024
value your f*cking time
September 23, 2024
the beauty of simplicity: why less is always more
September 2, 2024
what jodhpur taught me about life
August 26, 2024
startup lessons i’m learning