← All Work The Story The Product The AI The Business Impact Let's talk →
0→1 Build Predictive AI B2B HealthTech Medikabazaar 2022 – 2024

How I turned 100 hospital interviews into a predictive AI — deployed across India and UAE

Pen-and-paper wireframes to a live AI forecasting platform — built end-to-end as the only PM.

Sole PM 0→1 in 8 months India + UAE
Read the story ↓
≈4 min read

The Problem

Indian hospitals run inventory on gut feel.

Expired inventory — capital written off, compliance risk created.

Dead & non-moving stock — capital locked with zero visibility.

Gut-feel procurement — expensive last-minute sourcing, no intelligence.

MedikaCloud was built to be the intelligence layer none of them had — real time, SKU by SKU.

The Research

Before a single line of code, I met 100 hospital owners across India.

0+
Hospital owners
0
Cities
0
Users to design for
DelhiNoidaRewariMumbaiPuneKolkataVaranasiRaipurBangaloreHyderabad

The insight that shaped everything · The Buyer ≠ The User

The CXO buys

Driven by ROI & savings — wants the ₹ leakage view, never a spreadsheet.

The procurement manager uses

Driven by daily operations — indenting, vendor follow-ups, GRN.

"One product, two entry points — the ROI-seeker and the daily operator. Designing for this duality became the winning strategy."

The Product

Paper sketches → live command center

Hover any screen for the call behind it.

Home Dashboard
wireframe-home.png
Insight

"Lead with savings — it turns the app from an expense into a revenue-protecting tool."

Home Dashboard · the CXO's view

AI Forecast View
wireframe-forecast.png
Insight

"Not abstract graphs — a list of what to buy TODAY."

AI Forecast · prediction → action

Current Inventory
wireframe-inventory.png
Insight

"Procurement managers act without switching systems — every SKU, every day."

Current Inventory · the daily view

Expired & Near-to-Expiry
wireframe-expired.png
Insight

"Batch-level drill-down with cost attached — the problem hospitals never saw until too late."

Expired & Near-Expiry · invisible cost, visible

Create PO
wireframe-create-po.png
User pull

"UAE hospitals had zero legacy systems — we pivoted from 'only forecast' to 'full procurement' because users asked."

Create PO · user pull, not roadmap

Data Upload
wireframe-data-upload.png
Insight

"File-upload onboarding + GRN tracking — the operational backbone."

Data Upload & GRN · onboarding backbone

The AI

The model behind the product

Hospital demand isn't seasonal — it's irregular and life-critical. That's what made it hard:

Steady e.g. gloves, saline

Predictable. Classic forecasting works.

Lumpy e.g. implants

Quiet, then a spike. Smoothing models miss exactly what matters.

Erratic e.g. emergency drugs

No pattern at all. Needs per-SKU logic.

Three iterations to get it right

Prophet

Great at seasonality — smoothed away the lumpy spikes that mattered.

XGBoost

Caught non-linear patterns — couldn't scale across 1,500+ SKUs.

LightGBM

Fast and accurate on sparse, high-cardinality demand. Shipped.

What the model sees → what the hospital gets

Demand type

Steady / Lumpy / Erratic — different logic per pattern

Criticality tier

Life-saving SKUs get zero stockout tolerance

Geo layer

Weather + regional procedure trends — coastal ≠ North ≠ South

Medikabazaar's multi-year data

Pre-trained before day one — the moat no competitor could copy

LightGBM

per-SKU forecast

"Order today" list

What to buy, before the stockout

Finishes-in countdown

Days of stock left, per SKU

₹ leakage & savings view

Expired, dead and over-stocked capital, priced

The Business

One pricing line that did three jobs

The integration fee hospitals expected to be a blocker came back as procurement wallet credit.

The wallet flywheel

Integration fee 100% wallet credit → feels like ₹0 hospital procures via Medikabazaar → GMV ↑ intelligence + procurement in one ecosystem → lock-in switching costs compound every month
India · Medikabazaar customers India · Mid-market hospitals UAE · Independent clinics
SaaS Premium
₹49,999 / year
  • Unlimited SKU forecasts + AI indents
  • Full command center + PO/GRN module
  • 20%+ guaranteed savings on procurement
  • 100% wallet credit on integration fee
MedikaCloud landing page
medikacloud.com landing page

The calls that shaped it

Sole PM — every call without a safety net. All four paid off.

The model learned regional demand from multi-year data before it ever saw a hospital record — a head-start no competitor could replicate quickly.

Friction down, GMV up, lock-in compounding — three outcomes from one line in the pricing doc.

The CXO saw the ₹ savings; the procurement manager saw the SKU table. Same product, two entry points, two wins.

UAE hospitals had no procurement system, so I built the PO/GRN module because they asked. It drove 2 of our 5 paid conversions.

The Impact

8 months. 12 hospitals. Real numbers.

0%
Reduction in procurement costs
0%
Faster fulfillment turnaround
0%
Pilots converted to paid within 3 months
0
Pilot hospitals onboarded, India + UAE

Research to build to GTM to onboarding — one PM, one thread, never dropped.

Next Case Study

SensAI Sleep App

₹4.5Cr+ revenue in 12 months with AI sleep tracking

Read next →