Founder Message
Hey Product Hunt đź‘‹
I’m Fahad — first-year AI & Data Science student at UCLan and founder of Khwand.
Khwand started from a frustration I kept seeing while building with AI: shipping an LLM app or agent is easier than ever, but keeping it reliable in production is still incredibly hard.
You can build something that works today, only to have a model update, prompt drift, or edge case quietly break behavior tomorrow. And unlike traditional software, the failures are semantic — things don’t always crash, they just start behaving differently. Usually, you only find out when users complain.
That made me question something:
How do you actually ship AI-native software with confidence?
At first, I thought the answer was better testing. But while building Khwand, my thinking evolved. Static tests weren’t enough for probabilistic systems — AI apps need something closer to an autonomous reliability layer that can stress-test changes, catch regressions, and fix issues before users notice.
That shift became the foundation for Khwand.
Today, Khwand installs as a GitHub App and helps teams test and stabilize AI-powered code automatically:
→ Parses Python functions
→ Generates adversarial edge-case tests
→ Runs evaluations in sandboxes
→ Computes a Stability Score
→ Opens PRs with fixes when failures are detected
I’m very early (and building this while studying full-time), so this launch is really about learning in public, meeting builders, and validating whether this problem resonates as much as I think it does.
Three things I’d genuinely love feedback on:
Is “Stability Score” intuitive, or should it be called something else?
Which integrations matter most after GitHub? (GitLab, Bitbucket, CI tools?)
Are you building AI systems in Python, or would JS/TS support meaningfully change whether you’d use something like this?
Ask me anything — especially hard questions. I’m here to learn 🙌
Replies