Jigyasa Grover

Jigyasa Grover 12x AI + Open Source Award Winner • Google Advisory Board Member • LinkedIn Instructor • Book Author

I’ve been craving 鼎泰豐 Din Tai Fung’s chocolate xiao long bao for weeks now. Every time I checked, I just could not find ...
09/11/2026

I’ve been craving 鼎泰豐 Din Tai Fung’s chocolate xiao long bao for weeks now.

Every time I checked, I just could not find an evening slot 🫠

So the tired me, handed this task to Wajo AI’s Fo over iMessage, and I am impressed!

A handful of reasons …

→ Persistent, not one-shot. It didn’t just search once and give up, it polled availability at regular intervals, holding state across a task that spanned days, not a single session.

→ Guardrails instead of a leash. I gave it a payment method (a Visa debit gift card), so there was a hard ceiling on exposure, plus also configured hard spend limits. It gave me assurance to let it act autonomously instead of pinging me for a green light every step.

→ Recovered from failure instead of surfacing it as an error. The final confirmation hit a bot-check wall and got rejected. Most agents would’ve stopped and reported “failed.” This one retried through an alternate path in the background and kept me posted, without needing me to re-prompt it.

→ Kept optimizing after “done.” Reservation locked, and it’s still watching for a better (weekend) slot to swap me into - no fee, no extra ask, just continuing the objective. Big fan.

Isn’t this the real edge an agent has over a chatbot with tools? Holding a goal over time, recovering from failure, and knowing when to ask vs. just act.

I LOVE how Fo is acting like my digital EA.

It knows the objective and the boundaries, and figures out the rest across channels, over time, without needing me to babysit every step.

Well, who’s joining me for Din Tai Fung dinner? 🥟

Swipe to read the tale of how I found my favorite coding agent’s favorite agent 🤭LGTM
09/09/2026

Swipe to read the tale of how I found my favorite coding agent’s favorite agent 🤭

LGTM

08/04/2026

We put AI in the driver’s seat - literally 🏎️

Collaborated with Google Antigravity and Google Developer Experts to build something wild: an AI Race Coach whispering real-time tips into your ear at 100+ mph.

Pixel 10 + 100 car sensors + Gemma 4 running locally = zero cloud, zero lag, all thrill.

A few days ago it was just an idea. Then I was in the paddock, laptop melting in a 100° cockpit, cars screaming past, building live.

Trust isn’t prompted. It’s architected ⚡

This is what AI looks like when it leaves the cloud and hits the track.

h/t to for pulling this off!

Full video: https://youtu.be/SzmfH4P00vI

Every AI agent gives main character energy in the demo 💅  Then production hits... and suddenly there’s a plot twist nobo...
07/14/2026

Every AI agent gives main character energy in the demo 💅 

Then production hits... and suddenly there’s a plot twist nobody asked for. That’s exactly the gap I built my NEW COURSE to help close.

𝗘𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗻𝗴 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁𝘀 𝘄𝗶𝘁𝗵 𝗚𝗼𝗼𝗴𝗹𝗲 𝗔𝗗𝗞 is now live on LinkedIn!

It’s the follow-up to my previous course, “Building Agents with the Google Agent Development kit (ADK)” which 25,000+ of you have used to build your first agents.

Building an agent is step one. Trusting it in production is a completely different skill.

In this 2h 25m course, you’ll learn how to:
→ Design evaluation-ready agents with structured tool interfaces, Pydantic schemas, and reusable ADK templates
→ Inspect and audit agent reasoning using trace viewers, trajectory matching, and Golden Trace baselines
→ Measure reliability with headless batch evaluations, Pass@k testing, and statistical benchmarks
→ Build production safeguards with LLM-as-a-Judge evaluations, groundedness checks, guardrails, and CI/CD regression testing

Everything is hands-on: from GitHub Codespaces labs and live CLI demos to real-world failure scenarios and debugging workflows.

If you’re building AI agents and want confidence they’ll perform beyond the demo, I’m sharing limited-time FREE access, swipe till the end for the link.

Google Google for Developers

sneak peek into my notes from the “LLMs for Recommendation Systems” private event hosted by 📝
06/10/2026

sneak peek into my notes from the “LLMs for Recommendation Systems” private event hosted by 📝

05/30/2026

Two weeks ago, I ran a live no-code AI workshop at HQ in collaboration with for CSOs, physical security executives, insider risk leaders, and convergence practitioners.

1 hour. All hands-on. Build mode only.

By the end, everyone had an agentic planning workflow that:
→ asks before it assumes
→ separates facts from inferences from invented gaps, visibly
→ retains organizational context across sessions
→ keeps a full audit trail
→ shapes output for the right audience, without softening the evidence

Watching the same plan evolve across iterations. Same inputs. Same goal. Each version more capable than the last, and at every stage, showing exactly what the model made up versus what was actually grounded.

The reaction wasn’t panic. It was recognition.

“You gave me a lot to think about.”
“I learned a lot about agentic design patterns.”

Here’s the problem most teams don’t see until it’s too late: an AI-generated decision and an AI-generated guess look identical.

That’s not a model problem. That’s a design problem. And design problems have design solutions.

Reliability is not magic; it is engineered 🛠️

Address

San Francisco, CA

Website

Alerts

Be the first to know and let us send you an email when Jigyasa Grover posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Jigyasa Grover:

Shortcuts

Share

Category