Hi, I'm Shrey
I build AI agents that learn and evolve. Transitioning from test architecture to AI development taught me that reliability is key. I spend my mornings studying because this field moves fast, and I enjoy keeping up.
Working towards an AI-native future where agents don't just assist—they self-improve, self-heal, and handle the heavy lifting while humans focus on what matters.
The AI-Native Future
5 Years Outlook
AI agents will likely handle most coding tasks, with humans reviewing the output to ensure quality and alignment.
We'll see more agents that can self-improve and self-heal. I've been experimenting with agents that evolve autonomously based on user feedback and evals.
The engineering role is shifting towards orchestrating agents and designing guardrails. Understanding context is becoming as important as writing syntax.
Reliability in AI is achievable through evaluation-driven development. It's about bringing the discipline of testing to the flexibility of LLMs.
This is the future I'm excited to help build.
Things I Believe
Reliability in AI
A common concern I hear is that AI 'hallucinates' and can't be trusted. I've found that by taking a scientific approach—using evaluation-driven development—we can build much more reliable workflows. It's about testing and validating, just like any other software.
Context Engineering
In my workshops, I often see people trying to use one model for everything. I try to emphasize that providing the right context and guardrails is crucial. 'Context engineering'—structuring the information the AI receives—makes a huge difference in the quality of the output.
The Lightbulb Moment
I love that moment in workshops when people see a complex workflow come together. Whether it's building a full-stack prototype or a voice agent, seeing it work in real-time helps demystify the technology and shows what's possible.
The #1 Misconception
Many worry that AI will replace developers. In my experience, it's actually an amplifier. The engineers who embrace these tools aren't writing less code—they're building more ambitious systems and solving harder problems.
Quick links
