Mohammed Roshan Ibrahim A Available for work

I turn raw signal into decisions.

Applied AI and automation engineer. I was on the founding team of a DPIIT-recognised IoT startup, then built production AI agents and automation for about 28 client workstreams at Wyka. Based in Dubai.

  • Recognised by the Tamil Nadu Health Minister
  • Winner, IIT Madras Innowah
  • Two patents filed

Water quality, read continuously

WaterWise

Autonomous water-quality monitoring boat

A sensor suite on an ESP32 variant measures pH, turbidity, TDS and temperature while the boat moves. The hull and mounts were modelled in Fusion 360.

It is deployed at Madras Boat Club, it won IIT Madras Innowah, and the Tamil Nadu Health Minister publicly recognised the deployment.

Context an agent can act on

CAIRN

Memory and context layer for AI coding agents

Written in Rust as a Model Context Protocol server and running live at ccairn.club. It cuts LLM token use by 76.6% while keeping context coherent across sessions.

A judgment-gate review layer has caught real security issues in production use. Developers outside the project use it publicly.

Fifteen streams, one watcher

Hermes

Agentic monitoring at Hyrowah Technologies

Hermes is an agentic monitoring system built on open-source LLMs. It watches 15+ live IoT sensor streams and flags anomalies in under five seconds.

I built it as a founding team member at Hyrowah Technologies, a DPIIT-recognised startup, from May 2025 to April 2026. In the same period I owned Tank Pulse, a hardware and software product, end to end through a Chennai rollout to roughly 80 live field units.

Traffic to a lead worth calling

Everyone's Tomorrow

Growth platform, built end to end

Razorpay payments, segmented WhatsApp and email campaigns, Meta Ads, and Google Analytics attribution wired into one funnel.

The marketing side and the engineering side, built by the same person.

Steps that hand off on their own

n8n workflows

Self-hosted automation for clients and startups

I build n8n pipelines that connect databases, APIs, WhatsApp Business, Razorpay, Google and Microsoft 365 to LLM and vector-store nodes. Failing runs get retries, a fallback path and a human review step.

  • Copilot Studio agents Teams, SharePoint, Outlook; about 28 client workstreams at Wyka
  • Citizen-support RAG assistant n8n, Ollama, pgvector
  • Client onboarding WhatsApp, Razorpay, Sheets; about 70% less manual work
  • Anomaly-to-ticket MQTT stream to LLM summary to Jira
  • Executive report Postgres KPIs, written up by Claude, filed in Notion
  • Email triage agent Gmail, LLM, human approval before send

Where the work happened

Experience

  1. Apr 2026 — Sep 2026 AI engineer, Wyka, Dubai

    Six-month fixed contract. Built and ran production AI agents and automation on Microsoft 365 for about 28 client workstreams: Copilot Studio agents in Teams, SharePoint and Outlook, RAG over client documents with citations, and Claude and OpenAI calls wired into Power Automate and n8n. Tracked latency, token spend and quota, with retries and provider failover. Ran each engagement from discovery to handover.

  2. May 2025 — Apr 2026 Founding team member & AI engineer, Hyrowah Technologies LLP

    IoT and AI product delivery: firmware coordination, stream ingestion, backend APIs, inference pipelines, n8n anomaly-to-ticket flows and operator dashboards, across roughly 80 live units.

  3. Jan 2025 — Apr 2025 ML engineer intern, Pixel Expert Technology and Services

    Applied machine learning work across data preparation, modelling and evaluation.

Open to founding engineer and applied AI roles

Send a signal

Dubai, UAE. This page runs a GPU particle simulation — every point you scrolled through was solved on the graphics card, not animated by hand.