I build the backend for systems that move money.
Four years across insurance and capital markets. I led the team behind an insurance Point of Sale platform and wired up 200+ carrier APIs — now I'm building trading infrastructure at wizzer.in.
Tearsheet
Market-data streaming and the backtesting engine behind an AI quant & algo trading terminal.
drag it · click to turn it over
Opinions held with real money behind them, in the sense that I have had to live with each of these. Arguable on purpose.
Boring infrastructure
Postgres, a queue and one well-understood cache will outlive three clever services.
Columnar stores for time series
ClickHouse for ticks and bars, Postgres for everything else. Using one engine for both is how query times quietly become seconds.
Typed contracts at the boundary
Validate where data enters the system. Trust inside it, never at the edge.
Backpressure over bigger buffers
A feed you cannot keep up with is a design problem. A larger buffer only moves the failure later and makes it harder to read.
One adapter per partner
200 carrier APIs taught me this. A branch per quirk is a codebase nobody can change; an adapter per partner is one nobody has to.
Costs in the backtest
Strategies that only work at zero fees are not strategies. Same instinct as load-testing with realistic payloads.
Microservices before product-market fit
Distributed systems problems are a tax. Pay it when the scale is real, not when the org chart is aspirational.
ORMs in the hot path
Fine for CRUD, a liability where latency matters. Write the query; you will have to read it at 2am anyway.
Dashboards nobody opens
A chart that has never changed a decision is a maintenance burden with good branding.
Retry loops without a budget
Retries that do not count, back off and give up are how one slow dependency takes down everything in front of it.
Rewrites
The second system is slower to arrive and rarely better. Strangle it instead.
A practical guide to the two most important charts every algo trading dashboard needs — the Equity Curve and Trade Distribution Histogram — with React + Recharts implementations.
Learn how to format the success and error response in a standard format in Elysia.js middleware to ensure consistency across the application, handle the errors gracefully, and provide the user with the necessary information.
Create email service in Nest.js using MJML for responsive email template and handlebars for dynamic email content, and send email using Sendgrid (Twilio).