Haris Zahid, Senior mobile & full-stack engineer

Portrait of Haris Zahid

Monograph · Edition MMXXVI

Lahore, Pakistan

Senior mobile & full-stack engineer

Flutter apps, voice AI and the FastAPI backends behind them. I design, build, ship and run the whole product.

Move across the portrait to see the signal underneath

Taking on a small number of products this quarter

§ I · AI-assisted engineering

TheWorkshop

I build with AI agents every day, and treat them like a very fast team that never gets the final word.

The agents write. The machine verifies. I sign.

  1. 01SpecIntent written down before any code: the problem, the edge cases, what done looks like.
  2. 02AgentsIndependent tasks dispatched to coding agents in parallel.
  3. 03VerifyTests, builds and exit codes. A claim of success is not success.
  4. 04ReviewEvery diff read by me. Architecture and judgement stay human.
  5. 05ShipCI, beta tracks, then production, on servers I run.
  • Agentic coding

    Claude Code

    Daily driver for multi-file changes, refactors and reviews, and the engine inside Meridian, my agent orchestrator.

  • Agentic coding

    Codex

    A second agent for parallel tasks and an independent take on tricky diffs.

  • Agent-first IDE

    Antigravity

    Exploring and prototyping with agents that plan, edit and test across the workspace.

  • Model Context Protocol

    MCP servers

    Connecting agents to real systems: browsers, design files, docs and issue trackers.

  • Orchestration

    Agent swarms

    Dependency-aware waves of agents with shared memory, verified by the machine, not by their own reports.

  • In production

    Voice & API agents

    The same discipline in shipped products: Vapi outbound callers, LiveKit voice ordering, tool-using assistants.

4+ yrs
Building cross-platform apps in production
~30
Live projects on servers I run
iOS · Android · Web
Shipped to both app stores
1 owner
Design, build, deploy, operate

§ II · Selected work

TheVolumes

Eight products, each with the problem, my role, the stack and the result. Open any one for the full story.

More case studies

  1. The Quiet InfrastructureEngineer of record for production infrastructure · Zylo Technologies · 2025 — PresentAround 30 production projects on two servers, with backups that depended on someone remembering.LinuxDockerNginxPostgreSQLDiagram: Around 30 production projects on two servers, with backups that depended on someone remembering.
  2. The Watchful RoadMobile application engineer · Posh Rideshare · via Etechflow · 2025A US rideshare platform needed live driver and passenger monitoring on iOS and Android.FlutterKotlinSwift.NETDiagram: A US rideshare platform needed live driver and passenger monitoring on iOS and Android.
  3. The Dispatcher's VoiceBackend engineer · Dispatch product · via Zylo Technologies · 2026An outbound voice agent had to call at campaign scale without dropping calls.PythonFastAPIRedisVapiDiagram: An outbound voice agent had to call at campaign scale without dropping calls.
  4. The Rewriting EngineEnd-to-end engineer · MagicFeedPro · 2025 — 2026Rewriting thousands of product listings with AI risked damaging the whole catalogue.FastAPIPostgreSQLLLM pipelinesReactDiagram: Rewriting thousands of product listings with AI risked damaging the whole catalogue.

§ 0 A working thesis · revised 2026

The next generation of software will not have buttons. It will have brains and voices. And it will be judged by what survives production, not by what impresses in a demo.

Diagram: a phone under a spotlight on a pedestal, beside the same phone in rain, noise and weak signal, still showing a live orange voice waveform.
Fig. 1The demo is the easy half. Surviving the real world is the work.

Most products don’t fail in the demo. They fail in the gaps between the people who each built one slice. I close those gaps by owning the whole product, from the first sketch to the server it runs on, with AI agents doing the heavy lifting and nothing shipping that I haven’t verified.

01 — Mandate
Idea to income. One author, every layer, one name on the line
02 — Method
Augmented authorship: AI agents build, the machine verifies, I sign off
03 — Horizon
Products that listen, speak and act on their own
FlutterLiveKitWebRTCFastAPIKotlinLangGraphONNXPostgreSQLRedisDockerSwiftWebSockets

How I work · The Doctrine

Success that cannot be repeated is luck. This is the method.

§ III · Experience

Chronicle

From a single screen, to the whole product, to the servers beneath it.

Diagram: a single phone, then a cluster of devices, then a full system of devices, servers and databases, connected by a rising path.
Fig. 2From one screen, to the whole product, to the servers it runs on.
  1. 2026 — Present

    Nextreks

    Mobile Application Engineer · Freelance · Remote

    Takes features from raw design to production on iOS and Android; built CI on GitHub Actions so every commit builds and tests itself; runs beta programmes through TestFlight and Play Console; rebuilds signup, onboarding, and verification journeys to reduce drop-off.

  2. 2025 — Present

    Zylo Technologies

    Senior Mobile Engineer · Infrastructure · Full-time · Lahore

    Builds mobile apps, voice agents, and backend services for client products. Architect of Brewly's real-time voice ordering, lead on Scale App, backend for an outbound voice dispatch product, and engineer of record for two production servers carrying around thirty projects.

  3. 2025

    Etechflow

    Mobile Application Engineer · Full-time · Lahore

    Built real-time driver and passenger monitoring for Posh Rideshare, a US mobility platform, across Flutter, Kotlin, Swift, and .NET, and shipped releases to both app stores.

  4. 2025

    Tier75

    Voice Agent Engineer · Contract · Remote

    Co-built an always-on Android voice agent: hotword activation, LangChain and LangGraph tooling, AssemblyAI and on-device speech, ElevenLabs voice, Kotlin background services over WebRTC, and live web search.

  5. 2023 — 2025

    Independent practice

    Mobile & Automation Engineer · Freelance

    Mobile applications, games in Python, and automation in Bash and VBScript: the formative years of learning that every system is ultimately a pipeline.

§ IV · Tech stack

Instruments

Five strata, one owner.

Diagram: five stacked layers for intent, interface, signal, intelligence and infrastructure, pierced by one orange line.
Fig. 3Five layers, and one person who owns all of them.

Stratum 01

Intent

Why it should exist at all

  • Scoping & roadmaps
  • Figma to production
  • Onboarding & signup flows
  • Beta programmes
  • Release strategy

Stratum 02

Interface

Where people meet the product

  • Flutter
  • Dart
  • React Native
  • React
  • Kotlin
  • Swift

Stratum 03

Signal

Where milliseconds decide trust

  • LiveKit
  • WebRTC
  • WebSockets
  • openWakeWord · ONNX
  • AssemblyAI
  • ElevenLabs

Stratum 04

Intelligence

AI workflows that reason and act

  • AI agents via model APIs
  • Voice agents · Vapi
  • Voice agents · LiveKit
  • Tool calling & function routing
  • LangChain · LangGraph
  • LLM pipelines
  • Tavily live search

Stratum 05

Infrastructure

Where it all has to keep running

  • Python · FastAPI
  • PostgreSQL · Redis
  • Linux · Docker · Nginx
  • GitHub Actions
  • Backblaze B2
  • TestFlight · Play Console

§ V · Writing

Dispatches

Notes from the field.

§ VI · Contact · Commission

Remote · worldwide

When the demo works and production doesn’t.

Would twenty minutes on your product be a bad idea?

“No, not yet” is a perfectly good answer too.

Haris

Before you ask

  1. “One person can’t own all of that.”

    Agents share the labour. I own the outcome.

  2. “AI-written code is a liability.”

    Unverified code is. Mine is tested, then read.

  3. “Remote means chasing updates.”

    You get builds to tap, not status reports.

  4. “I’ve been burned by a freelancer before.”

    That’s why the method is published.

Languages · English for work · also Urdu, Punjabi and Pashto