Bhaskar Thota← All work

Case study · Client product · Built end to end

Mark My Zone

A production property management platform that replaced a real estate firm's spreadsheets, then learned to predict repairs before they happen.

ProductPredictive analyticsFull stack, AI assistedNext.js · Supabase
170+Properties in production
50%Operational time saved
9Repair types forecast
0→100Risk score per flat
01 · The story

A 170 flat portfolio, run out of spreadsheets

A real estate firm was running an entire rental portfolio of 170+ flats out of spreadsheets. Rent, leases, tenants, owner payouts and repairs all lived in separate sheets that only one or two people fully understood.

Lease renewals slipped. Rent went uncollected because nobody had a single view of who had paid. Repairs were always reactive, a call came in only after something had already broken.

So I sat with the people doing the work, scoped what actually slowed them down, and built the whole product end to end on Next.js, TypeScript and Supabase.

02 · The product tour

The portfolio explains itself

Below the headline numbers, the data tells its own story. Collection trends, the occupancy mix, and the repair categories quietly eating the budget.

Every empty flat shows the revenue it is losing while it sits vacant. The information was always there. This finally makes it legible.

03 · The product tour

Analytics anyone can read

A dedicated analytics view turns months of operations into trends, not tables. Punctuality per flat, collection over time, and where maintenance spend really lands.

It is built for the owner who wants the health of the portfolio at a glance, without asking anyone to pull a report.

04 · The predictive engine

Then it starts predicting

This is where it stops describing the past. The engine turns the firm's own repair history into a forecast, and a single risk score from 0 to 100 for every flat.

Each category has a rhythm. AC roughly every six months, plumbing closer to nine, a geyser about a year. Once a flat has its own history, the engine learns that flat's real interval and adapts.

At a glance: how many flats are at risk, what is overdue, and the cost likely to land in the next ninety days.

05 · The predictive engine

Down to the single flat

Open any flat and it lays out the per category forecast. What is overdue and by how many days, what it is likely to cost, and the history behind the call.

Reactive maintenance becomes a planned schedule. The team fixes the riskiest units first, before a tenant ever has to call.

06 · The result

Reactive became proactive

The spreadsheets are gone. The team runs the whole portfolio from one dashboard, and the work that used to eat their week is roughly 50% faster.

The predictive layer changed the business from reactive to proactive. Instead of waiting for a breakdown, the team sees which flats are at risk and what the next ninety days will cost.

It shipped to production and runs as the firm's internal operations tool. Because it holds private tenant and owner data, the screens here use representative demo data rather than real records.

Tech & tools

Next.jsTypeScriptSupabase (Postgres, Auth, RLS)RechartsPWAVercelPDF & Excel exportClaude Code (AI assisted build)

My role

  • Ran the discovery and user research with the client, and scoped requirements into a buildable product.
  • Owned the product end to end: data model, feature set, UX, and delivery to production.
  • Designed and built the dashboard as an AI assisted full stack build on Next.js and Supabase.
  • Conceived and implemented the predictive maintenance engine: the interval learning, risk scoring and forecasting logic.