I built a self hosted runtime for routing decisions that learns from feedback

Аналитика рынка

Стек

Откликнуться

Загрузите резюме — мы свяжем вас с работодателем напрямую через нашу базу.

Описание

I built a self hosted runtime for routing decisions that learns from feedback Hey all! I have been working on a project called Syntra. It started because I got tired of hardcoding routing logic and manually tweaking it every time traffic changed or a provider started acting weird. A while back, when AWS had issues, our traffic basically stopped, which obviously is not acceptable. That pushed me toward building something that could route between different nodes or cloud providers automatically based on latency, health, and anomalies. I also tried it for LLM routing as an experiment, and it worked surprisingly well. The idea is pretty simple. Syntra sits in front of repeated operational choices and helps decide what action to take. Then it learns from feedback over time and keeps an audit trail of what it chose and why. I am currently using it inside our API stack to pick the best route for traffic in real time and detect anomalies within milliseconds. The obvious use cases are node routing, LLM model routing, provider selection, retry policies, autoscaling decisions, fraud thresholds, and similar scenarios where the best answer depends on live context. I also made some fun demos for it, including Mars launch-window planning using NASA/JPL data, an epidemic policy simulator, and an edge-of-chaos discovery demo. [link] Would genuinely like feedback, especially from people who have had to build routing or decision logic like this in production. [link] [handle]

Контакты работодателя (email/phone/telegram) скрыты из публичного превью — отправьте резюме, чтобы мы связали вас напрямую.