01
AI opportunity audit
I find where AI actually matters in your business (a cost, a bottleneck, a decision made on stale data) and put a number on it before anyone commits budget.
VIREVOL CONSULTING SERVICES
24 years building AI in production,
mostly before there was a name for it
When I tell a client what to build, I've already built it: the harness, the evals, the serving layer, all of it. That's what happens when most of your career predates the tools you'd otherwise buy.
Email me directlyHOW WE CAN HELP
Evidence before production spend
An engagement is scoped around finding a real cost or opportunity, proving the fix, then shipping it. Not a strategy deck.
02
Shadow-mode proof
I build the system and run it against your real operations without touching production. You see the projected outcome before you spend a dollar deploying it.
03
Production deployment
I ship the model, the eval harness, the serving layer, and the monitoring into production myself, with a small team on delivery, then hand it off in a state your team can run.
WHY THIS IS DIFFERENT
Built before it had a name
Most AI consultants started in the last two or three years. This is what being early actually looks like, dated.
2013
DeepDressAI: deep-learning dress recognition on Torch Lua at Rent The Runway, before "deep learning" was a hiring line.
2013
A simulation framework gating every recommendation-algorithm change at RTR against hundreds of test cases: evals, before "evals" was a word.
2023
Logos Shift, open-sourced: a proxy that fine-tunes and auto-rolls cheaper models into production, a year before Braintrust and similar products existed.
2024
Tora, my own agent harness: agentic chat-to-video with closed-loop performance editing, live before coding agents went mainstream.
2023
Bohita named a Microsoft launch partner for the Bing/ChatGPT plugin ecosystem, unveiled at Satya Nadella’s Build 2023 keynote: the world’s first ChatGPT-to-physical-product experience.
THE PATTERN
Find the unfunded bet.
Prove it yourself
Three dated instances of the same move, fifteen years apart.
2010Barnes & Noble
Hired as a contractor to do analysis. The work justified building a data science team from scratch, and I became its first hire. Recommendation work from that team ended up on receipts in every store.
2012Rent The Runway
Hired to "personalize" against a SQLite database and text files, when the CEO didn’t believe fashion recommendations could work at all: it was a bet by the CDO. I stood up a Vertica data warehouse on my own initiative, proved it could handle the load, then got headcount to build the team around it. When deep learning needed GPUs nobody would fund, I paid for AWS out of my own pocket to prove visual search worked before the company adopted it.
2025SGShipping
Same move, most recently: found bunker fuel as the real lever, proved it in shadow mode, then earned production deployment.
CASE STUDY: SGSHIPPING, SINGAPORE
Found the cost. Proved the fix. Shipped it
Engaged to define an autonomous vessel strategy. I audited revenue and cost drivers first and found bunker fuel: the single largest addressable cost, and a number directly measurable enough to make it a fair place to prove an agentic system before touching production.
I built a vessel-routing system, ran it in shadow mode against live operations, and reported predicted versus actual savings daily. Leadership asked for deployment once the evidence was in front of them, not before.
Result: an agentic system matching vessels to routes now saving 30% of fuel in production. Alongside it, a Level 1–5 autonomy roadmap built with their engineers, maritime operators, and regulators.
Same pattern at Bohita Ads: agents read live campaign performance on Meta, Google, and TikTok and edit themselves. Full campaign in ~10 minutes, 97% average watch rate on Meta.
30%fuel savings in production, proven in shadow mode firstFROM PEOPLE WHO HIRED ME
Testimonials
“Saurabh helped us go from a demo to a product. He has the ability to go up and down the stack and worked with me closely on product. Sundial was deploying multiple ML models on the edge every day. Highly recommend.”
“Thank you Saurabh! We would not be here without you! Thank you for all you have done to build RTR and our special culture. Always here to support you and so proud of you!”
“I appreciate the work, passion and ideas. Glad you got us into the GPT Store. We would have been unlikely to do that without you.”
WHO THIS IS FOR
Operating businesses in retail,
fashion, logistics, and legal
I've founded and run three AI companies myself (Bohita, Virevol, TidyLaw), owning product, engineering, and P&L, not just the recommendation.
I lead every engagement and remain accountable for the technical and business decisions. My small remote team supports implementation, testing, integration, and ongoing support.
An initial conversation is a chance to find the one place in your business where this is worth trying, and to agree on what evidence would change your mind before we start.
Rent The RunwayMembership program built from scratch, 50%+ of revenue
Barnes & NobleFirst predictive analytics team
NVIDIAInception program, spoke on AI at the edge at GTC
TidyLawPatent research, adopted by a top law firm
Shopify merchantsRecommendation and pricing tooling
SGShippingVessel routing, 30% fuel savings
START HERE
What's your highest-leverage cost?
Send me the P&L line you'd most want to cut. I'll reply within 48 hours with one of three things: it isn't provable, it's provable but not worth doing, or here's exactly what I'd measure to prove it.