Selected work

What we have built, and how we build.

Systems we have designed and built, written up as patterns: the problem, our approach, and what it changes.

Capability patterns

Solution patterns for demanding operational environments.

Each pattern is a system we have built. Client names are left out; the problems, approach and stack are real.

Capability pattern

InsureTrack: policy expiry and follow-up management

Problem: Agents tracked hundreds of policies in spreadsheets. Renewals slipped, follow-ups lived in memory, and bulk data arrived as messy Excel files with duplicates and missing phone numbers.
Approach: One system for customers, policies and follow-ups. Expiries are ranked Critical to Low by days left, follow-ups carry a timeline and history, and an Excel import maps columns, validates rows and flags duplicates. A data-quality view lists records missing critical fields.
Impact: Every policy in one searchable book, renewals worked from a priority list instead of a calendar, and imports that clean the data on the way in.

Priority-ranked expiry alerts
Excel import with validation
Follow-up timelines
Insurance
Product Engineering

Stack

Next.js
TypeScript
ASP.NET Core
SQL Server
Clean Architecture

Capability pattern

Guest-house CRM and property management

Problem: Owners ran several properties on registers, OTA dashboards and WhatsApp. Double bookings, unbilled extras and late follow-ups were routine, and compliance paperwork for foreign guests was done by hand.
Approach: A multi-property system covering reservations, room calendar, check-in and check-out, housekeeping, maintenance and GST invoicing. Two-way OTA sync, a public booking site with Razorpay prepayment, rule-based WhatsApp and email automations, owner approvals and role-based access.
Impact: Double bookings blocked at the database, bills generated at check-out with the right tax split, and guest messages sent by rules instead of reminders.

Two-way OTA sync
GST invoicing per property
Event-driven automations
Hospitality
Product Engineering

Stack

React
Node.js
PostgreSQL
Razorpay
Gemini API

Capability pattern

Smart retail POS for furnishing stores

Problem: Busy shops billed by hand, staff gave away margin while bargaining, customer credit sat in a notebook, and nobody knew the true day's profit until month end.
Approach: Barcode billing that works with any scanner, a minimum-price lock with owner PIN override, split cash and UPI payments, and WhatsApp receipts. Udhaar ledgers for customers and suppliers, an evening cash-drawer audit, multi-store stock transfers, and a room-colour visualiser for showroom TVs. Fully bilingual in Hindi and English.
Impact: Faster counters, no sales below break-even without the owner's approval, credit tracked with automatic reminders, and a real net profit every evening.

3-second barcode billing
Margin-protecting price lock
Hindi and English UI
Retail
Web & Mobile

Stack

Next.js
TypeScript
Prisma
PostgreSQL

Capability pattern

HRMS for growing teams

Problem: Employee records, attendance and leave were spread across spreadsheets, chat threads and paper. Approvals took days and monthly reports were rebuilt by hand.
Approach: One HR system with employee profiles, attendance and leave workflows, manager approvals and exportable reports, with separate access for HR, managers and staff.
Impact: Requests approved in one place, a single source of truth for people data, and reports generated rather than assembled.

Leave and attendance workflows
Manager approvals
Role-based access
HR & Operations
Product Engineering

Capability pattern

Conversational assistants on WhatsApp and the web

Problem: Teams answered the same questions all day on WhatsApp and email, and leads that arrived after hours waited until morning.
Approach: Assistants grounded in the business's own content, available on WhatsApp and the website, with hand-off to a person when a question needs one and a log of every conversation. The assistant on this site is one of them.
Impact: Routine questions answered instantly at any hour, and people freed for the conversations that need them.

Grounded answers
Human hand-off
Conversation logs
AI Integration
Customer support

Stack

Gemini API
Claude API
WhatsApp Business API
Python

Capability pattern

Healthcare Journey CRM

Problem: Patient messages, bookings and records sit in separate systems. Staff retype details, appointments get missed, and nobody sees the whole journey.
Approach: One portal for messaging, scheduling and record access, with role-based permissions and an audit trail on every read. Each clinic keeps its existing record system; we integrate rather than replace.
Impact: A single patient timeline, fewer missed appointments, and intake that takes minutes instead of a phone call.

Consent-aware data sync
Scheduling assistant
Cross-channel notifications
Product Engineering
Cloud & Security

Stack

Next.js
PostgreSQL
FHIR APIs
AWS
WhatsApp Business API

Capability pattern

Multi-Channel Communication Platform

Problem: Every product team rebuilds the same email, SMS and WhatsApp sending, with different templates and no shared view of what was delivered.
Approach: A single notification service with versioned templates, per-channel rate limits and delivery events. Teams send one event; the service picks the channel and records the outcome.
Impact: One place to see deliverability, no template drift between teams, and automatic fallback when an SMS gateway fails.

Template versioning
Rate limiting
Delivery telemetry
DevOps & Automation
API Engineering

Stack

Node.js
Kafka
Redis
Twilio
SendGrid
Grafana

Capability pattern

Multi-Cloud Operations Portal

Problem: Workloads span AWS and a private Kubernetes cluster. Access, cost and alerts live in different consoles, so problems are found late.
Approach: A single operations portal that pulls metrics, access logs and cost data from every environment. Policy checks run automatically and flag drift before it becomes an incident.
Impact: No blind spots across environments, one set of policies enforced everywhere, and cloud spend visible by team.

Unified access federation
Cost allocation dashboards
Automated policy checks
Cloud Infrastructure
Observability

Stack

Terraform
Kubernetes
Prometheus
Grafana
OpenTelemetry
AWS

Capability pattern

Intelligent Workflow Automation

Problem: Hundreds of unstructured customer requests arrive each day. People read each one, pull out the details, and route it by hand.
Approach: A language-model pipeline extracts the structured fields, checks them against the knowledge base with vector search, and routes the request. Anything uncertain goes to a human, and every decision is logged.
Impact: Routine requests are classified automatically, people handle only the unusual ones, and every automated decision can be reviewed.

Structured extraction
Human-in-the-loop review
Audit logging
AI Integration
Product Engineering

Stack

Claude API
pgvector
LangChain
Python
PostgreSQL

Start with the challenge

Ready to improve what technology can do?

Tell us what you need to build, modernise, automate, or scale. We’ll help define the most practical path forward.

Start a conversation

No obligation. Just a useful first conversation.