Chief Technology Officer · Kakushin.io LTD

I run engineering for a 39-project software portfolio.

CTO of a Cyprus-registered software company, accountable for architecture, delivery, infrastructure and security across 15 internal products and 24 client engagements — web, native mobile and desktop — with a distributed team of around twenty across two time zones.

Before that, a decade of real-time communications engineering: WebRTC telemedicine, remote-control streaming protocols, and cross-platform apps shipped from Islamabad, Kuala Lumpur and Silicon Valley startups.

Asghar Ali, Chief Technology Officer at Kakushin.io LTD
Chaudhary Asghar Ali · Cyprus · Pakistan

Scope

What sits under the role.

39
Projects under technical ownership

15 internal · 24 client

~20
People in the organisation

12 engineers, plus design and QA

45
Repositories migrated

To company-owned source control

30+
Live production domains

Across the estate

Selected work

Four of the thirty-nine, chosen for what they demanded rather than what they earned.

Regulated · under subcontract

National electricity utility — customer apps

Dual native iOS and Android customer applications for a European national electricity utility, delivered under subcontract. 60+ screens over 8 backend services, dual-brand builds, and a fortnightly two-platform release cadence.

Coordinator and escalation owner across seven external engineering counterparties. Drove third-party penetration-test remediation gating production acceptance, and introduced a relay-proxy architecture that removed every client API key from the shipped binaries. Custom EN/EL localisation and certificate pinning.

  • /SwiftUI
  • /Kotlin · Compose
  • /Keycloak
  • /Azure APIM
  • /Oracle CC&B

Multi-vendor marketplace

BookMi

Visit

Booking and events marketplace spanning venues, packages and transport, with web, admin and React Native surfaces. The most service-dependent product in the portfolio.

Production-freeze governance on main while go-live was withheld; the merge gate on daily client-feedback releases; escrow and payout change control; OTP-path risk analysis. Established a 233-test Playwright suite against 148 documented use cases.

  • /Node.js
  • /React Native
  • /MongoDB Atlas
  • /Stripe
  • /Twilio

Healthcare

Dementia Buddy

Visit

Consumer dementia-support app plus a care-provider portal with a twelve-module caregiver training curriculum. Five distinct account types, and a BLE wearable module with proximity, panic alerts and GPS breadcrumbs.

Access-hygiene review on a healthcare dataset, PII handling review, LLM cost attribution and endpoint-abuse analysis, and App Store submission through to approval.

  • /Next.js
  • /React Native
  • /Firestore
  • /MongoDB
  • /OpenAI

Internal product · live revenue

AI Humanizer

Visit

AI-text detection and humanisation SaaS — the portfolio product carrying real subscription revenue. Full admin console covering users, subscriptions, credits, abuse and fraud, marketing, email and system health.

Runs its own email service with self-managed DKIM and delivery tracking, rather than depending on a third-party sender.

  • /Vercel
  • /Firestore
  • /OpenAI
  • /Stripe
  • /Custom SMTP

Built personally

The tooling the engineering function runs on.

CTO Dashboard

An Electron and SQLite desktop application joining Slack, GitHub and the in-house tracker: live presence, hours reconciled against what was claimed, scrum compliance, commit scan across the portfolio, and cross-project release tracking.

Automated PR pipeline

Sweep → review → fix → merge → deploy → announce → release, driven by the Claude Agent SDK behind a code-level preflight that refuses draft PRs, protected branches and secret-shaped diffs.

Dashboard MCP server

The same database exposed to any LLM session as twelve read-only tools, each answer carrying a freshness block and the caveats a reader would otherwise get wrong.

SprintForge

An in-house Jira replacement — projects, sprints, issues, releases, boards — adopted company-wide as the system of record, and exposed as an MCP server for LLM-driven workflows.

History

Ten years, mostly spent on things that had to work in real time.

  1. Present

    Chief Technology Officer

    Kakushin.io LTD · Cyprus

    Technical direction across a 39-project portfolio; a distributed team of ~20 over two time zones; infrastructure, security posture, release governance and vendor strategy.

  2. 2021 — 2026

    Lead Developer

    SwotKnot · Boulder, CO

    Learning platform matching students to tutors by learning style using in-house machine-learning models. React front end, Node.js services.

  3. 2020 — 2021

    Lead Developer & Software Architect

    RYZA · Palo Alto, CA

    Real-time co-creation platform allowing multiple people to remote-control software from the browser. Designed the architecture around WebRTC and a proprietary protocol for streaming application control over a data channel.

  4. 2018 — 2020

    Backend Developer

    PAO · Boulder, CO

    Node.js services, deployments and releases for a local-discovery media marketplace.

  5. 2017 — 2018

    Cross-Platform App Developer

    Open Dynamics · Selangor, Malaysia

    Mobile, Angular front ends and Java back ends for Malaysian government software, including four shipped Play Store applications.

  6. 2015 — 2017

    Full-Stack & Telemedicine Engineer

    CASE University · Islamabad

    Built a WebRTC clinical research tool for remote patient examination, and real-time chat and video conferencing for the telemedicine department. Part of the team behind AL-BASR, winner of a 2015 P@SHA ICT Award. Lead engineer on a wearable ECG device.

Stack

What I actually work in.

Languages & frameworks

  • TypeScript
  • Node.js
  • React
  • Next.js
  • React Native
  • Electron
  • Swift / SwiftUI
  • Kotlin / Compose
  • Python

Real-time & media

  • WebRTC
  • SRTP / DTLS
  • TURN / coturn
  • FFmpeg
  • HLS
  • MediaMTX

Data

  • MongoDB
  • Firestore
  • PostgreSQL
  • Redis
  • SQLite

Cloud & infrastructure

  • Vercel
  • Cloudflare
  • Hetzner
  • Render
  • Firebase
  • nginx / PM2

AI

  • OpenAI API
  • Anthropic Claude
  • Claude Agent SDK
  • MCP server authoring

Practice

  • Architecture review
  • Release governance
  • Playwright
  • GitHub Actions
  • Pen-test remediation

Research

Where the computation should sit, and how to secure it.

Fog & edge computing

My final-year work implemented a fog computing architecture in Node.js, including placement of the fog layer inside an SDN environment. The premise — push computation toward where data is produced rather than hauling it to a distant cloud — is the same one that now governs how real-time and inference workloads get placed.

  • /Fog architecture
  • /SDN placement
  • /Latency budgets

Edge AI

Running inference next to the sensor rather than in a datacentre: model partitioning between device and edge node, the accuracy traded for latency and power, and the operational reality of shipping model updates to hardware you cannot SSH into.

  • /On-device inference
  • /Model partitioning
  • /Update pipelines

AI & cyber security

Two directions at once. AI as a defensive instrument — anomaly detection over traffic and behaviour. And AI as new attack surface: prompt injection, tool-call abuse, model and data exfiltration, and the supply-chain question of what a generated dependency actually contains.

  • /Prompt injection
  • /Tool-call abuse
  • /Anomaly detection

Real-time communication

The oldest thread and still the deepest: WebRTC telemedicine from 2015, remote-control streaming protocols, and the transport problems — NAT traversal, relay economics, congestion control — that never stopped being interesting.

  • /WebRTC
  • /Telemedicine
  • /Transport

Also

I run a real-time communications practice.

Alongside the CTO role I run Daniotech, a WebRTC and real-time communication practice trading as Danio LLC — building video calling into products, migrating teams off metered SDKs onto self-hosted LiveKit and mediasoup, and fixing the calls that fail for a subset of users and cannot be reproduced.

It is also where the deeper WebRTC writing lives — TURN and coturn setup, ICE debugging, SFU topology.

Contact

Open to conversations about engineering leadership, real-time systems, and portfolios that have outgrown their process.