Abhishek’s Portfolio Open to work

Case study 02 · Public website

Real Estate Platformpublic website

A Delhi brokerage’s property site, built around how people search: type a sentence, check a colony’s price band, or ask Roofie, a scripted enquiry assistant. Relaunched in Sep 2026 alongside its new self-hosted CRM.

6,564

active listings on the site

6 Nov 2026

20

organic listing enquiries

9 Sep → 6 Nov 2026

26 → 103

engaged visitors from Google search a week

W38 → W44 · 14 Sep – 1 Nov 2026

Status
Live at rentaroof.in since the 9 Sep 2026 relaunch
My role
Product owner, AI-assisted build. I set the requirements and made the product decisions; AI coding tools write the code under my direction, and I test every change. Earlier versions of the site were built by outside teams I coordinated.
Built with
Laravel · Livewire · MySQL, one codebase with the CRM
Integrations
Meta Pixel and Meta Conversions API · Google Search Console API · email · Knowlarity IVR and DLT SMS through the shared CRM
Roofie
A scripted, rule-based enquiry assistant, not AI
Search results on desktop for "3 bhk in vasant vihar under 1.5 lakh": 3 BHK listings in Vasant Vihar with the Roofie panel beside them
FIG. 01Search, desktop: one typed sentence, read as filters, with Roofie open beside the results. Demo copy, fictional data.
Roofie chat on a phone: a visitor types a 3 BHK brief for Vasant Vihar and Roofie replies with the top live matches
FIG. 02Roofie on a phone: a typed brief becomes matches from live stock.

01 · In short: What’s there? What’s fair? Who do I talk to?

Renters arrive with the same three questions. The brokerage was relaunching its website alongside its new CRM, and visitors don’t think in dropdowns. They type a sentence, want to know what’s fair, and want a person when they ask for one, even after hours.

02 · What it does

  • Plain-English search

    One box takes a whole sentence; bedrooms, colony and budget become filters you can remove.

  • Price guide

    Rent and sale bands, with gross yield, by colony; a market report shows its working.

  • Roofie

    A scripted assistant that shows matches, quotes a price band, saves a search, requests a visit, takes an enquiry and hands over to a person.

  • Listing pages that say what was checked

    When each fact was verified, the building’s other units, and a sticky Enquire bar on phones.

  • Compare and shortlists

    Up to four listings side by side, plus advisor shortlists shared as one link.

  • Customer dashboard

    Signed in with a phone number and a one-time code, no password; saved homes, enquiries, visits and new matches.

  • List your property

    Owners submit a property on the site; the desk checks it before it’s listed.

  • Locality guides and journal

    Written guides for 77 colonies, and a journal of blog and news posts, 37 published (6 Nov 2026).

03 · What makes it different

  1. Read plain English. “3 bhk in vasant vihar under 1.5 lakh” becomes a Vasant Vihar chip plus bedroom and budget filters, each one removable.
  2. Filtered lists are the landing pages. A search such as “3 BHK for rent in Vasant Vihar” is itself the landing page, with its own title and Google tags, so there is no second set of pages to maintain. 599 such locality × type pages sit among the 7,474 sitemap URLs built from live stock.

    6 Nov 2026

  3. Show what’s fair. Rent and sale bands, with gross yield, by colony, from a pricing matrix the team maintains, not an average of asking prices.
  4. Scripted by design. Roofie is a scripted, rule-based enquiry assistant, not AI: no external API and no LLM, as the brokerage asked. Any thread can pass to a person.
  5. Channel ≠ capture. A visitor who googled, then used Roofie, reads “Google search” + “Roofie”, so search gets credit for the lead.
  6. One copy of every listing. The site reads straight from the CRM’s inventory, with a publish switch on each listing, so there’s nothing to keep in sync, and a home that’s been let or sold can’t be published.

04 · What the business can configure: Run from the CRM, without a developer.

Listings
Each listing has a publish switch in the CRM; publish refuses a let or sold home.
Homepage
The headline, photo and stat bar are edited in the CRM.
Search shorthand
“DC” for Defence Colony is a synonym row admins edit in the CRM, with no code change.
Roofie’s answers
Paperwork and viewing questions are answered from the FAQ the team edits in the CRM, so editing an answer changes Roofie’s reply.
Locality guides
Guides the team writes in the CRM (notes, nearby places and photos), 77 on 6 Nov 2026; map, prices and live listings are added automatically.
Journal
Staff draft, schedule or publish, with each post’s category, cover and featured flag.
The rest of the site
Homepage, new pages from ready-made blocks, menus, footer and FAQ, under separate permissions.

05 · Key screens

Screens: white-labelled demo copy, fictional data, unless marked live.

01Search

Search page with a Vasant Vihar colony chip, removable filters and matching 3 BHK listings
FIG. 03The colony chip, applied filters you can clear, then the matching homes.
The same search on a phone: one search bar, a filters control and listing cards
FIG. 04The same search on a phone, filters folded into a single control.
Home page search box with example searches to tap underneath
FIG. 05Home: one box, worked examples to tap, and classic filters one link away. The headline, photo and stat bar are edited in the CRM.

02Pages from live stock

Landing page for 3 BHK homes in Chittaranjan Park, noting few matches and showing nearby listings
FIG. 06A locality × type landing page with few matches says so, then shows nearby stock. Demo stock.

03Price guide

Price guide page: market headline figures and colony cards with rent and sale bands
FIG. 07Price guide: headline market figures, then one card per colony with rent band, sale band and gross yield. Demo figures.
The price guide on a phone: headline figures and the first colony card
FIG. 08The price guide on a phone. Demo figures.

04Roofie

Roofie opened on a listing: the listing card, an enquiry sent, and buttons to check availability, request details, WhatsApp or call
FIG. 09On a listing, Roofie opens on the home being viewed and offers availability, details, WhatsApp or a call. Demo data.
The same Roofie chat in the CRM inbox on a phone, with a Take over button
FIG. 10In the CRM inbox, the desk takes over the same chat in one tap. Demo data.

05Listings & account

Property page on desktop: photo gallery, price, verified details and the Roofie panel
FIG. 11Property page: gallery, Enquire / Save / Share, the details the team verified on the ground, and Roofie already on this listing. Demo data.
The same property page on a phone: price, a 'Verified by our team' badge and a sticky Enquire bar
FIG. 12The same listing on a phone: price, “Verified by our team” and a sticky Enquire bar.
Compare page on desktop: two 3 BHK listings side by side with configuration, built-up area, floor, furnishing and locality rows, each with an Enquire button
FIG. 13Compare: up to four listings side by side, with a Floor row, each an Enquire tap away. Demo data.
The customer dashboard on a phone: tiles for saved properties, open enquiries, visits booked and new matches
FIG. 14Customer dashboard on a phone: saved homes, open enquiries, visits and new matches, read from the person’s CRM record. Demo account.

06 · Results, dated

Found on Google within a month.

Engaged visitors from Google search per week

First-party tracker · IST weeks · part weeks pale · 11 Sep – 6 Nov 2026

Engaged visitors from Google search per week, first-party tracker, IST weeks 37 to 45 of 2026; week 37 (11–13 Sep) and week 45 (2–6 Nov) are part weeks
WeekCount
W37 (part week)3
W3826
W3934
W4041
W4189
W4277
W4379
W44103
W45 (part week)47
  • ~27,000

    Google search impressions in 23 days, about 2.7× the daily rate of the first month after the relaunch (Search Console, includes brand searches)

    14 Oct – 5 Nov 2026

  • 26 → 103

    engaged visitors from Google search a week, within two months of the relaunch

    W38 → W44 · 14 Sep – 1 Nov 2026

  • Page 1

    on Google for rent searches in 5 key South Delhi localities, within 30 days of the relaunch (Search Console + manual checks)

    Sep–Oct 2026

  • 499

    engaged visitors from Google search since 11 Sep, averaging 5 min 35 s each; 82 landed on listing pages

    11 Sep – 6 Nov 2026

  • 43%

    of engaged visitors came from Google search (375 of 880), up from 22% in the 30 days to 23 Oct

    30 days to 6 Nov 2026

  • 6,467

    site searches since 11 Sep; searches naming a colony also feed the verification desk’s list of where to check next

    To 6 Nov 2026

  • 32

    website & Roofie leads (22 Roofie property enquiries, 8 Roofie chats, 2 website forms); 30 organic, 12 from Google search; 17 of 32 (53%) now qualified

    9 Sep – 6 Nov 2026

  • 20

    organic enquiries about a specific listing, all through Roofie on the listing page; 11 arrived from Google search

    9 Sep – 6 Nov 2026

  • 101 of 119

    Roofie chats never needed an agent (85%); 35 became or joined an enquiry

    7 Sep – 6 Nov 2026

  • 7,474

    URLs in the sitemap: 6,561 listing pages, 599 locality × type landing pages, 163 hubs, 77 locality guides, 41 agency pages, 33 journal & site pages

    6 Nov 2026

  • 11

    owner submissions since the form went live: 6 listed, 4 dropped, 1 new; a working supply channel, not a growth claim

    20 Sep – 6 Nov 2026

Counted by the site’s own tracker (live since 11 Sep 2026), which filters known bots, burst crawlers and signed-in staff; Google figures count engaged visitors only. Total visitor, pageview and raw Google visitor counts are not quoted, because automated traffic still inflates them. Lead figures come from the CRM’s read-only query of 6 Nov 2026, about 2 months after go-live (9 Sep – 6 Nov): across every channel, 147 of 496 new leads were organic by source (not from paid ads or portals; 145 not linked to any ad), and only the 30 organic website and Roofie leads (12 from Google search) are website-attributable.

07 · How it was built: Same codebase as the CRM, built AI-assisted.

  1. Requirements and product decisions.I gathered what the desk and visitors needed and made the calls.
  2. Built with AI coding tools.I directed the tools and tested every change before it shipped.
  3. Shipped with the CRM.Site and self-hosted CRM replaced a 12-seat subscription CRM in 3 months (Jun–Sep 2026); production cutover 9 Sep 2026. I ran go-live and training.
  4. Before this version.Earlier versions of the site were built by outside teams I coordinated.

StackLaravel · Blade + Livewire · MySQL · same codebase as the CRM