See
Perception & diagnostics
See what the business actually does — the patterns, the inconsistencies, and the work that repeats.
An engineering atelier for intelligent systems
We engineer intelligence into the way things work.
Avisha creates AI systems that can see, analyse, reason, detect, explain and act inside real-world operations — built around the problem, the domain and the people who use them.
Intelligence, thoughtfully applied.
reads the system as it is01 / 03
Services
Purpose-built intelligence for real systems. From perception and diagnostics to reasoning, decision support and intelligent execution.
Think
Reasoning & decision intelligence
Analyse context, compare possibilities, explain findings and strengthen human judgement.
Act
Intelligent execution
Take appropriate actions, coordinate tools and respond to changing conditions within designed boundaries.
Build
Specialised AI systems
Custom applications and intelligent products shaped around the domain, the problem and the people who rely on them.
Work
Selected systems, tools and experiments we have built to explore how intelligence can improve real work.
Operating system
AIOS
A company operating system that keeps work moving between decisions. Five operating primitives coordinate the flow; four fixed laws protect it. A founder steps in only when judgment is genuinely required.
Stack →
Operating system
AIOS
ClaudeNext.jsSupabaseVercelGoogle Workspace
Autonomous delivery
The Build Pipeline
An AI-assisted delivery system that turns a defined brief into a working application. Reusable templates and specialist skills guide the build, while an independent review layer tests the work before it reaches a human.
Stack →
Autonomous delivery
The Build Pipeline
Claude CodeGitHubSupabase MCPVercel
Stock market intelligence
Signal to Order
Signals from charts, Telegram, WhatsApp and other sources are evaluated against the portfolio and market trend. Strategies can be backtested and paper-traded before orders reach the market—on a human tap or through an autonomous trading bot.
Stack →
Stock market intelligence
Signal to Order
Next.jsVercelPython/FastAPIRailwaySupabaseClaudeKite ConnectTelegramWhatsApp/TwilioTradingView
Inventory control
Stock & Movement
Successive stock counts reveal what has sold, what is moving quickly and what is slowing down—even when a sales report is missed. Every movement is tracked, so replenishment and recall decisions follow the true demand for each design.
Stack →
Inventory control
Stock & Movement
Next.jsTypeScriptTailwindSupabaseRechartspdf.jsExcelJSVercel
Private by design
BetterUS · Shared Health Record
A shared health space that brings together data from each partner’s wrist, phone and connected health apps. Couples can encourage healthier choices and gently remind one another to move more, while selected details remain private.
Stack →
Private by design
BetterUS · Shared Health Record
ViteReactTypeScriptSupabaseVercelApple HealthWHOOP API
Workflow automation
Operations Dashboard
Enquiries and operational data from every channel are captured, validated and unified. Clear cases move automatically, uncertain ones go to the model, and notifications keep people informed. The weekly position then compiles and sends itself.
Stack →
Workflow automation
Operations Dashboard
Claude CodeSupabaseGoogle SheetsTelegramWhatsApp
CRM intelligence
Lead Qualification & Follow-up
A CRM and outreach system that finds and qualifies relevant organisations, verifies contact details and removes duplicates. Personalised messages are paced deliberately, while every follow-up and response is written back to the record.
Stack →
CRM intelligence
Lead Qualification & Follow-up
n8nApifyOpenAIAnymail FinderGoogle SheetsSMTP
Accounts & records
Sales Desk
An accounting application that brings sales invoices, sales returns, collections, stock and packing slips into one properly managed record. Built-in OCR reads photographed documents and turns paper into structured data, while records can be organised, reviewed and removed when no longer required.
Stack →
Accounts & records
Sales Desk
ViteReactTypeScriptTailwindshadcn/uiSupabaseNetlify
Process
Engineering before implementation. We start with the system, identify where intelligence can genuinely add value, and build only what earns its place.
ObserveStep 01
Observe
01 / 05We sit with the system before we touch it — the data, the people, the constraints, and the decisions that actually get made.
Find the opportunity
02 / 05We look for the places where better perception, reasoning or action changes the outcome — and what it would take to get there.
Design & Build
03 / 05We define how the system should behave — its boundaries, its inputs, what happens when things go wrong, and what a person sees and when. Then we build it against real data, from the first day.
Verify
04 / 05We test it against what it will meet in the world — edge cases, bad data, the days that don't go to plan — and tune it until it holds.
Hand over
05 / 05Built for you. Belongs to you. We build for independence, not dependency. Your system should remain understandable and operable after delivery.
Founder
I was trained to see systems before I was trained to see software.
Avisha was founded by Mrinal Gala, a Chemical Engineer with a Master's degree from the United States, years of experience on large industrial projects, and a granted patent.
That background sets the standard here. We analyse the system as a whole — how work moves through it, where it slows, what it could do if it were more intelligent — and then engineer for that, at whatever scale the problem actually is.
FAQ
What types of businesses does Avisha work with?
Any industry. The problem matters more to us than the label on it — if a system produces data, involves judgement and has people depending on it, we can work with it.
How long does a typical engagement take?
It depends on what we find. A focused system can be live in weeks; a larger one takes longer. We agree the scope before we build, so you know what you're getting and when.
Do we need technical expertise on our team?
No. We design for the people who will actually use the system. What we do need is someone who knows the work well enough to tell us how it really happens. Documentation and handover are part of the build.
What happens to our data?
Your data stays yours. We work inside your systems where possible, we don't train anything on your data, and access ends when the engagement does.
What happens after the project is delivered?
You own the work — source access, documentation and the system itself. No lock-in. Ongoing support is available if you want it, not required.
What does it cost?
Every system is scoped and priced individually, after we understand the problem. We'll tell you what something will cost before any work starts.
How do we start?
Start by showing us the problem, system or decision you want to make more intelligent. We look at the context first, then define what kind of build — if any — would genuinely add value.