Hi, I'm Ishan. I turn strategy and data into products, and chaos into consensus.
I love building cool stuff with cool people.
I started in data and never really left. I just kept trading up for bigger problems to solve, which pulled me from SQL and dashboards into Python and machine learning, and eventually into a seat where I had a say in what we built and who we brought in to ship it. These days I'm building AI agents and products that make sense of payments risk at Walmart. AI is a great equalizer: it knocked down the wall between having an idea and actually shipping it.
Lately that same curiosity has me poking under the hood, into how systems scale and stay secure, learning each new piece as the work demands it. The constant across all of it: build things people love, not just things that work.
I'm an introvert who learned to love a full room of people. When I'm not building, you'll find me on a trail, in a video game, watching anime, or lost in a book. I also take food and coffee far too seriously. And I'll happily talk your ear off about any of it.
Multi-Agent AI Credit-Risk System · Walmart
I handed a team of AI agents a job I used to grind through by hand: dig through credit risk, surface what actually matters, and recommend a move. They run on their own now, and have already flagged dozens of high-risk sellers before things got expensive.
Production ML Risk-Scoring Model · Walmart
Led a risk-scoring model from blank page to production, shoulder to shoulder with data scientists and a lot of AI. It catches high-risk sellers about 77% of the time, and its calls now drive real decisions: credit limits, funding, suspensions.
AuditBoard (Optro) Product Ownership · Delta
Owned Delta's rollout of AuditBoard (now Optro) across five functions, from Finance to Internal Audit. Automated 45 SOX control tests (there went 700+ hours of annual busywork) and shipped 33 analytics products that took reporting from weeks to nearly real time.
Fraud & Recoupment Analytics · Deloitte
Built the analytics that caught what people missed: payroll fraud, money worth clawing back, and compliance risk buried in large loan portfolios. The findings didn't sit in a deck, they shaped what leadership did next.
Agentic Trading
An ongoing exploration of how agentic systems can reason about markets, weigh ideas, and act with a human in the loop. Mostly an excuse to learn how autonomous agents actually behave when the stakes feel real.
TasteGraph
An AI app for people who take food too seriously (guilty). It learns your palate, the cuisines, flavors, and vibe you actually go for, turns it into a food personality, then points you to restaurants worth your time.
SnoutSketch
A playful web app that turns a close-up of your dog's snout into one-of-a-kind art, reading the unique ridges of their nose and rendering it in styles from vibrant to surreal to line art. Dheera was the first model.
Ten years, three chapters, each building on the last: from data, to machine learning, to AI and product.
Beyond the day job, the ideas I happily fall down rabbit holes on, the ones that genuinely bring me joy.
AI Is a Fast Track to Average
The wall between having an idea and shipping it just came down. What it was hiding is the whole game.