Recent work
What I'm building — and the decisions behind it.
Live platforms where I own architecture and revenue calls, and products I'm designing and selling myself. Each case study shows the problem, how it works, the architecture decisions and the business model.
Case 01 · Architecture & Revenue Decisions
Dail-it
Local business discovery for India, run entirely from a phone.
One app where local businesses get a digital business card, an Instagram-synced catalogue, WhatsApp reach and home-service booking — and where customers nearby find, compare and contact them.
The problem
Most small businesses in Indian towns — salons, plumbers, clinics, coaching centres — have no website and no one to build one. Customers still find them by word of mouth. Websites, hosting and domains are a cost and a skill barrier they will never cross.
01How it works
- 1
Create a free listing
About five minutes of phone form-filling. No coding, domain or hosting.
- 2
Add services & photos
Hours, location, contact, and an Instagram-synced product gallery.
- 3
Customers search nearby
By name or category, compare ratings and details.
- 4
Call, WhatsApp or book
Enquiries and home-service visits land straight with the owner.
02Key features
Netbiz
Location-based discovery: businesses show up in nearby searches with hours, location and contact details.
InstaBiz
Syncs Instagram product posts into a shoppable gallery on the business card, so the catalogue maintains itself.
WhatsApp Anyone
Open a WhatsApp chat with any number without saving the contact — the channel customers already use.
Homebiz
Home-service booking: schedule visits, track appointments and keep customer history in one place.
03Architecture decisions
- 01
One platform, four products — not four apps
Discovery, catalogue, messaging and booking share one business profile. Each product adds value to the same listing instead of splitting the audience.
- 02
Free listings are the growth engine
Supply is the hard side of a local marketplace. Zero-cost listings fill every city first; upgrades monetise businesses once they are getting enquiries.
- 03
Reuse Instagram and WhatsApp instead of rebuilding them
Owners already post products on Instagram and talk to customers on WhatsApp. Syncing with them removes data entry and meets customers where they are.
- 04
Android-first, iOS path kept open
The target market is overwhelmingly Android. Shipping there first got real usage fast, with the iOS release planned on the same backend.
- 05
Infrastructure fully managed by the platform
No surprise bills, servers or domain renewals for the owner. Removing that fear is a big part of why they sign up at all.
04Revenue model
Listing
Free
Profile, discovery, contact buttons — builds supply in every city
UpgradesCore
Paid tiers
Catalogue, booking and visibility features for businesses already getting enquiries
Customers
Always free
Search, compare and contact — demand is never taxed
- Mobile (Android)
- Location search
- Instagram sync
- Booking engine
Case 02 · Client Project · Agency Website
AyudhSEO
A lead-generating website for a full-service digital marketing agency.
The web presence of an India-based digital marketing agency with offices in New Delhi and Chicago — SEO, social media, PPC, reputation management, content and web design, sold to businesses in the US, UK, Australia and India.
25+
Team members
100+
Accounts handled
12+
Countries served
98%
Client satisfaction
The problem
An SEO agency's own website is its sales pitch: it has to rank for the services it sells, prove results to sceptical buyers abroad, and turn visits into enquiries — while the marketing team keeps it updated without waiting on developers.
01How it works
- 1
Visitor lands from search
Service pages target intents like SEO services India or local SEO.
- 2
Compares packages
Each service links to its own package page — SEO, SMO, PPC, ORM, web design.
- 3
Checks the proof
SEO results, portfolio, client reviews and video testimonials.
- 4
Sends an enquiry
Contact forms and direct phone / email for a custom quote.
02Key features
Service landing pages
Dedicated pages for SEO, social media marketing, PPC, online reputation management, content writing, local SEO and web design.
Package pages
Separate SEO, SMO, PPC, ORM and website-design package pages, so buyers can self-qualify before they get in touch.
Proof library
SEO results, portfolio, our-work, client reviews and video testimonials — trust signals for overseas buyers.
SEO reseller channel
A reseller page offering white-label SEO to other agencies — a second revenue channel on the same site.
03Architecture decisions
- 01
WordPress with the Astra theme
The marketing team publishes and edits pages themselves. For a content-led agency site, editorial speed matters more than a custom stack.
- 02
One URL per service and per package
Each page targets one search intent, which is how an SEO agency's own site earns rankings for the exact services it sells.
- 03
Structured data via Yoast
Organization, WebSite, SearchAction and breadcrumb schema give search engines a clean picture of the business and its pages.
- 04
Proof pages as first-class navigation
Buyers in the US, UK and Australia hiring an offshore agency need evidence before a call, so results and testimonials get their own pages.
- 05
Quote-led conversion
Packages set expectations, but final pricing happens on a call — forms, phone and email are on every page to capture the lead.
04Engagement model
PackagesCore
Per service
Published SEO, SMO, PPC, ORM and web-design packages
Custom
On request
Scoped quotes for larger or multi-channel accounts
Reseller
White-label
SEO delivered under other agencies' brands
- WordPress
- Astra
- Yoast SEO
- Schema.org
- Contact Form 7
Case 03 · Own Product · Building in Public
AI-Drafted Reply Service
Customer email replies drafted in the business's own voice.
A done-for-you service for small businesses: every new customer email in Gmail gets a drafted reply built from the business's own FAQs, pricing and tone. The owner approves, edits or discards — nothing sends unsupervised.
$300–800
Setup fee
$100–250
Monthly retainer
10
Clients to first goal
~$2k
Target MRR
The problem
Owners and office managers at small businesses answer the same customer emails every day — prices, availability, policies. Generic AI tools sound generic and get details wrong, so they don't trust them near customers.
01How it works
- 1
New email arrives
An automation watches the client's Gmail inbox.
- 2
Sent with business context
Email text plus FAQs, pricing, policies and tone go to Claude.
- 3
Draft generated
Anything uncertain — disputes, complaints — is flagged, not guessed.
- 4
Owner reviews & sends
Approve, edit or discard from a review queue.
02Key features
Per-business knowledge base
FAQs, past replies, pricing, policies and tone-of-voice examples collected at onboarding and refined monthly.
Tightly scoped drafting prompt
The system prompt instructs the model to flag sensitive or unclear emails instead of inventing answers.
Human-in-the-loop queue
Drafts land in a shared doc, Slack channel or lightweight dashboard for approval before sending.
Selective auto-send (upgrade)
Once trust is built, low-risk categories like FAQ answers can send automatically; everything else still routes to a human.
03Architecture decisions
- 01
Draft, don't auto-send
Trust is the product. A review step protects the client's customers — and me — if a draft is ever wrong, and it makes the sale easier.
- 02
Wire existing tools, don't build a platform
Claude API plus Zapier / Make.com / Gmail API keeps the engineering lift small, so effort goes into each client's context and prompt quality.
- 03
One channel, one niche
Gmail only, and one niche (real estate, boutiques or local services), so the pitch, demo and prompt template are reused for every prospect.
- 04
Setup fee plus retainer, few clients
High revenue per client relative to effort. Ten clients on a retainer beat hundreds of low-paying self-serve subscribers.
- 05
Sell with a demo on real data
A short recording — email in, draft appears, owner sends — using one real business's FAQs closes deals; a description of the idea doesn't.
04Pricing
Setup
$300 – $800
One-time: connect the inbox and build the knowledge base
RetainerCore
$100 – $250 / mo
Upkeep and ongoing prompt and knowledge-base refinement
Target
≈ $2,000 / mo
10 clients at a $200 retainer
A working plan — pricing and timelines are starting points to adjust after real client conversations.
- Claude API
- Gmail API
- Zapier / Make.com
- Slack
- Prompt engineering
Case 04 · Own Product · Design Done, Shipping Next
AuditSnap
Catch tracking that fires before consent — and turn it into a sales audit.
A Chrome extension that detects pre-consent tracking — Meta, Google, TikTok, LinkedIn pixels and session recorders — and converts each finding into a 1-click, white-labelled PDF audit that web agencies use to sell compliance-fix retainers.
$49/mo
Agency ARPU
85%+
Gross margin
$58.8k
ARR at 100 agencies
< 4 wks
MVP build
The problem
Many sites fire ad pixels on page load before the visitor touches the cookie banner. Sending an IP or browser fingerprint to Meta or Google before consent is a GDPR / CCPA / CIPA exposure — even if the site stores no names or emails. Agencies need a fast, credible way to show this to prospects.
01How it works
- 1
Open the client's site
Public pages, staging, localhost or password-protected portals.
- 2
Intercept before consent
Network calls and cookies are captured before banner interaction.
- 3
Grade the risk
Each finding scored PASS, WARNING or CRITICAL.
- 4
Export a branded PDF
Agency logo, colours, disclaimer and call-to-action in one click.
02Key features
Network & storage interception
Chrome declarativeNetRequest and cookies APIs detect calls and cookies dropped before cookie-banner interaction.
Traffic-light severity engine
PASS: tracking correctly blocked. WARNING: analytics with incomplete privacy flags. CRITICAL: ad pixels or session recorders firing on load.
1-click white-label PDF
A branded audit with the agency's logo, palette, custom disclaimer and direct call-to-action — rendered client-side.
Staging & localhost scans
Runs inside the browser, so it inspects environments that external server crawlers can't reach.
03Architecture decisions
- 01
A browser extension, not a server crawler
Runs on the user's machine: near-zero server cost, sees staging and logged-in portals, and observes exactly what a real visitor's browser sends.
- 02
Evidence from raw network logs, not DOM guesses
Banner detection alone throws false positives. Proving an HTTP request actually reached an ad server is defensible in front of a client.
- 03
Built for Manifest V3 from day one
declarativeNetRequestWithHostAccess plus background service workers keeps it within Chrome Web Store rules instead of fighting them later.
- 04
Client-side PDF generation with jsPDF
No rendering servers to run or pay for, and audit data never leaves the browser.
- 05
Sell to agencies, not site owners
Agencies turn each audit into a $500–$2,000 fix retainer, so a $49/mo tool pays for itself with one closed deal — a low-CAC, recurring B2B model.
- 06
Tracker rules from open lists
EasyList and Disconnect give broad, maintained coverage of known trackers from the first build.
04Pricing tiers
Starter / Dev
$0 / mo
3 manual scans a month, on-screen report
Pro Developer
$19 / mo
Unlimited scans, CSV export, staging & localhost
Agency GrowthCore
$49 / mo
Unlimited scans, white-label PDF exports, 3 seats
Enterprise
$149 / mo
Unlimited seats, API access, bulk domain monitoring
At $49/mo: 25 agencies ≈ $14.7k ARR, 100 ≈ $58.8k, 250 ≈ $147k, 500 ≈ $294k.
05Risks & mitigations
False positives on consent banners
Mitigation: Findings rest on network request logs proving the call reached a third-party server.
Legal liability
Mitigation: Every PDF carries an automatic disclaimer: technical diagnostics, not legal advice.
Manifest V3 restrictions
Mitigation: Interceptor designed around declarativeNetRequestWithHostAccess and service workers.
- Chrome MV3
- React
- declarativeNetRequest
- jsPDF
- EasyList / Disconnect
Next step
Tell me what's breaking.
You don't need a spec, a budget approved, or a tidy summary. Send the symptoms and what you've already tried. That's enough for me to give you a useful answer.
- 01
You send the symptoms
Two paragraphs is plenty. No NDA needed to describe a slow API.
- 02
I reply within 24 hours
With my honest read on what's likely happening — free, no call required.
- 03
We scope it, or we don't
If it's a fit you get a fixed price and a start date. If not, I'll say so.