Software engineer at Square Sparx and The Hub, Mauritius

I build software that firms trust with their clients' data.

Compliance platforms, AI chatbots and business back offices. I design, build and run each one end to end.

Products I built and run

The StatutoryIQ website, showing the action queue, entity register and Outlook sync.

StatutoryIQ

Statutory compliance software for corporate service providers and company secretaries. One workspace replaces the spreadsheets and inbox threads.

In use at MINDEX.

statutoryiq.com

What I built

  • An entity register, recurring filing obligations and a daily action queue that ranks what is overdue, unpaid or coming up.
  • An Outlook add-in that files email attachments straight to the right client entity.
  • Role-based access, so staff see only the entities assigned to them, with a full audit history.
  • Now building an onboarding assistant that reads KYC documents on our own servers. No client document goes to an AI provider.

Stack

  • TypeScript
  • Bun
  • Hono
  • tRPC
  • Drizzle
  • PostgreSQL
  • S3
  • Next.js
The SparxBot website, with a live chat window answering a visitor.

SparxBot

A chatbot trained on your website. It answers from your own pages and links the source of every answer.

sparxbot.com

What I built

  • A pipeline that crawls a site, indexes it in Postgres with pgvector and answers only from what it finds. When it does not know, it says so and saves the question.
  • Domain-locked keys, rate limits and a daily spend cap for every site.
  • Workflows that let the bot call the owner's own API, such as booking a table. Visitors pick from buttons, confirm each action and verify their email before seeing private data.

Stack

  • Python
  • FastAPI
  • TanStack Start
  • PostgreSQL
  • pgvector
  • Redis
  • Docker

Built for clients

Brytland

Brytland Office

The internal platform Brytland uses to place candidates in university programmes.

  • Candidates sign up, upload documents against a checklist and apply to programmes.
  • Staff review documents, record payments, send applications and log each university's reply.
  • Reports on the pipeline, outcomes and team workload, with CSV export and a full audit log.
  • TanStack Start
  • tRPC
  • Drizzle
  • PostgreSQL
  • S3
  • Better Auth

Wholesale distribution

Distribution back office

The back office a distributor uses to supply the island's hypermarket chains.

  • Product catalogue with images, categories and stock movements.
  • Customer orders, invoices and a point of sale.
  • Sales reports on revenue, top products, top customers and order status.
  • Next.js
  • tRPC
  • Drizzle
  • PostgreSQL
  • Better Auth

How I work

The firms I build for handle passports, bank details and legal deadlines. These rules keep that work safe.

  1. Security comes first

    Tenant isolation, least privilege, audit logs and real deletion are designed in from the first commit, not added later.

  2. Plain code before AI

    If a rule can decide something, code decides it. Models get the jobs only they can do, and code checks their work.

  3. People make the judgement calls

    The software gathers the evidence. A named person approves, rates and signs off, and the history shows who did what.

  4. Client data stays home

    Sensitive documents are read on our own servers. Nothing reaches a third-party model without a data protection review.

  5. Small, reviewed steps

    Big features ship one slice at a time. Each slice is built, tested and reviewed before the next one starts.

How I use AI

I add AI where it does a job better than plain code, and I make sure its answers can be checked.

SparxBot

Grounded answers

Retrieval over a customer's own pages, with a source link on every answer and a polite refusal when the answer is not there.

risk-copilot prototype

Evidence-checked extraction

Drafts a client risk assessment from an onboarding pack. Every field cites a quote, and code confirms the quote really is on that page. It scored 46 out of 46 on six synthetic test packs.

StatutoryIQ

A private onboarding assistant

Local OCR and passport MRZ parsing, sanctions and PEP screening on our own servers, and a model that picks the next action but never writes case data.

Now learningAI engineering from first principles, with machine learning as the next step.

Tools I use every day

Languages
TypeScript, Python, SQL
Frontend
React, TanStack Start, Next.js, Astro, Tailwind CSS
Backend
Hono, tRPC, FastAPI, Better Auth
Data
PostgreSQL, pgvector, Drizzle, Redis, S3
AI
OpenAI API, Retrieval (RAG), Evals, OCR
Infrastructure
Docker, Coolify, Bun

Square Sparx and The Hub

I work for Square Sparx Ltd and its parent company, The Hub Business Solutions. Together we build our own products and custom software for businesses in Mauritius and the UK.

Square Sparx Ltd

Software company in Mauritius. Home of StatutoryIQ and the compliance work on this page.

The Hub Business Solutions

Parent company and software house, with offices in Mauritius and Birmingham, UK. Runs SparxBot.

thehubbusiness.com

What you can hire us for

  • Compliance and client onboarding software
  • AI chatbots and assistants on your own data
  • Back offices and internal tools
  • Hosting, support and ongoing improvements

Tell me what you need built.

A new product, an internal tool, or AI added to software you already run. I reply within a working day.