Abhishek
Srivastava

The part of engineering I actually care about is what happens after the demo works. Getting a model to generate one nice image is easy; keeping it reliable under real traffic is the real job — so I tend to think in latency budgets, failure modes, and guardrails before I think in features.

That shows up in how I build. I like capabilities that fail independently instead of taking the whole service down, config-driven systems where onboarding a new client is a database row and not a redeploy, and infrastructure engineered to run lean. I'd rather own a service end-to-end — schema, safety gates, deploy — and cut first-view latency from 29 seconds to 11 than ship a feature nobody asked for.

I'm most at home where backend rigor meets messy AI behavior, and I'm still hungry to go deeper — lately I've been sharpening on Go and distributed systems while finishing a master's in AI/ML.