Tumpreet Singh

Backend-focused full-stack engineer · AI/ML

I'm a backend-focused full-stack engineer who also does AI/ML. The through-line across everything I build: the real-time systems and the ML that make AI products actually work in production — plus the tooling around them.

Most of my work lives where latency, concurrency and reliability collide: streaming voice-AI pipelines, model serving with batching, multi-tenant middleware, and the observability and evaluation harnesses that keep all of it honest.

Background: an MCA foundation plus agency-hardened delivery — shipping real things for real deadlines, then going deeper into systems and ML. I care about the parts users never see and always feel.

How I work

01 Latency is a feature: If it feels slow it is broken. I budget it, measure it, and defend it.
02 Boring where it counts: Postgres, Redis, queues. Novelty belongs in the product, not the plumbing.
03 Idempotent by default: Assume at-least-once, out-of-order, and retried. Design so replays are safe.
04 Tooling is leverage: Eval harnesses, replay CLIs, dashboards — I build the thing that builds the thing.
05 Ship, then prove it: Metrics over adjectives. A number on a dashboard beats a claim in a deck.

→ SYSTEMS → USES → EMAIL