Fractional CTO · Technology & Product Leadership

Shrinath Balakrishnan

I turn business problems into working technology — and keep it lean enough that the tech budget never becomes the story.

Eight years as co-founder and CTO of Source.One, owning technology and product for one of India's largest B2B industrial distribution businesses.

Download CV
₹3,000 CrRevenue the platform carried, FY26
0.08%Tech spend as share of revenue
20–25Engineers running the whole platform
19% → 7%Freight cost error, after ML
< $1MCumulative platform cost, eight years

Source.One, FY2018–FY2026.

What I do

Selected work

Three decisions that shaped a ₹3,000 Cr business

Judgment under constraint

The platform I threw away in week one

Inherited an unsalvageable build with a live go-to-market deadline.

Source.One's first platform arrived from a subcontractor: an over-engineered microservices build with poor coding standards, and a product concept that forced buyers and suppliers through separate logins — an adoption killer in a market that transacts on WhatsApp.

The call: not recoverable inside a three-to-six month go-to-market window. So we didn't try.

Instead: WhatsApp as the front end, because it was already universal among buyers, suppliers and transporters. A CRM as an improvised back end. Custom middleware for messaging, algorithms and automation. Two-way conversational commerce, years before it was a category.

INHERITED MICROSERVICES WHATSAPP FRONT END MIDDLEWARE CRM AS BACK END CUSTOM PLATFORM BUYER + SUPPLIER APPS
MVP shipped inside the launch window, then migrated deliberately as each layer became the bottleneck.

We replaced the improvised back end once it became the constraint, then shipped native buyer and supplier apps for the users who'd outgrown a chat interface. Every migration happened when the business needed it — not before.

ML with a P&L outcome

Freight estimates that stopped eating the margin

Cost-estimate error cut from roughly 19% to under 7%, nationally.

In industrial distribution, freight is quoted before it's incurred. Every point of error between the estimate and the actual cost comes straight out of a margin that is thin by definition — and across a national network of lanes, vehicle types and seasons, the estimates were running roughly 19% out.

BEFORE ~19% AFTER <7% 0% 25%
Gap between estimated and actual freight cost, before and after the model.

Under 7% after the model shipped. That difference is margin recovered directly, on every movement, in a business where the whole game is unit economics.

It sat alongside a personalised minimum-price engine, real-time supplier price matching per order, and a reverse-auction freight quotation system that secured the lowest price for each vehicle movement.

Engineering leadership

Twenty-five engineers, then twenty

Restructured into three divisions and cut headcount while the business grew.

Most engineering stories are about growth. This one runs the other way: the platform behind a ₹3,000 Cr business was built and run by 20–25 people, and I deliberately trimmed it to 20 while revenue kept climbing.

That was possible because of how the function was organised:

PRODUCT ENGINEERS SDE + PM merged SOFTWARE ARCHITECTS depth track AI ENGINEERS standing division THREE SPECIALISED DIVISIONS
Product Engineers own outcomes end to end — the PM layer folded into engineering rather than sitting beside it.

Merging the SDE and PM roles meant engineers owned product outcomes rather than receiving tickets. A separate architect track kept technical depth from being flattened by delivery pressure. AI Engineers were a standing division, not a side project.

Delivery ran on explicit must-have / good-to-have / nice-to-have prioritisation, held firmly. That discipline is why go-to-market windows were makeable and why cumulative technology spend stayed under $1M — 0.08% of revenue.

Range

I've dropped into unfamiliar businesses before

Depth at one company matters less than evidence I can learn yours quickly. Before Source.One I spent two years in a consultancy doing exactly that — a new domain every few months, architecture chosen per problem rather than mandated.

Red Panda 2015–2017

Digitised Sodexo's paper meal-voucher system; built a Kenyan smart-energy IoT product zero-to-one; launched GlobalCitizen's India platform for the Coldplay concert with a two-person team.

Amazon 2012–2013

Retail Services' global ML-driven competitive pricing platform — billions of data points daily. Led an end-to-end AWS rewrite of a subsystem.

Microsoft 2011

Azure disaster-management team: analysed a multi-region cloud outage and designed mitigation strategies.

Education

B.Tech Computer Science, NIT Warangal. MS Aerodynamics & Computation, University of Southampton.

Next step

If the tech budget has become the story, that's usually a solvable problem.

Fractional CTO engagements, technology advisory, fixed-scope builds — and I'll consider the right full-time CTO seat.

Email me
CV

Happy to talk.

What's this about? It helps me come prepared.

Privacy

This site uses cookieless analytics that records page views, approximate country, device type and referring link. It sets no cookies, stores nothing on your device, and cannot identify you.

Resume links carry a short code identifying which application or channel they came from, so I know which listing generated a visit. It identifies the link, not you.

If you book a call or email me, the name, email and details you provide are used only to arrange and prepare for that conversation. To have them deleted, email nitarsh@gmail.com.