// senior backend engineer · 8 years · payments, data, AI agents
Data systems that
prove it.
I'm Muhammad Mairaj. I build data platforms and the AI agents on top of them, then measure whether they work.
See what I've built About me- 8 yearspayment and data systems
- NayaPaypayments for Pakistani merchants
- 3 blockchainssettlement and KYC at Normies
- 13 s → 0.8 sTrino lake lookups
- 5 npm packagesagent guardrails and tools
Integrated
Rain stablecoin cards Persona KYC
deBridge cross-chain swaps
Privy wallets and gas sponsorship
// answers only from my posts and projects · it can be wrong, the posts are the real thing
warehouse-brake
A Claude Code plugin that stops an agent before it spends Databricks or Snowflake credits, even through dbt.
az-brake
The same brake for Azure: the agent asks before any command that creates or starts a resource that costs money.
wrap-it-up
A per-repo memory for coding agents, so the next session starts without a recap.
data-lab2
One health-survey pipeline built on Databricks and Snowflake, with every gold value matched across both.
pageviews-on-a-laptop
143M rows a day of Wikimedia pageviews loaded on one laptop, with every row accounted for.
tick-ledger
Live exchange trades through Kafka and Spark into Delta Lake, every candle checked against the exchange.
What a load balancer does, explained simply
One address for your users, many servers behind it, and a quiet check that keeps broken ones out of the way.
System designVertical vs horizontal scaling, explained simply
A bigger machine or more machines. Both work, and the choice changes how you build.
System designCaching 101: where a cache sits and what it saves
A cache is a copy kept closer to where it is needed. Every layer you add skips some of the trip.
Data lakes and query enginesRegister Iceberg tables in a Postgres catalog for faster Trino queries
A small Postgres database, two tables and one row per Iceberg table took my Trino lookups from 13 seconds to under one.
EssaysExplaining a data platform decision to people who don’t write SQL
The numbers were right. The unit was wrong, and an analogy did more work than any chart.
Data lakes and query enginesTrino vs StarRocks vs Doris on an Azure Iceberg lake
Same tables, same machine, same six questions. One engine was fastest, one was most familiar, and one couldn’t read the files at all.