← OnTriggerJoyPortfolioResumes↓ PDF

Shahryar Saqib

I build AI agents that do real operational work, with a person approving anything risky.

Case study: https://ontriggerjoy.com/work/ai-automation/

Summary

I design multi-agent AI systems and put them to work on real tasks. My own fleet of AI bots runs my email, finances, records, job search and company operations from one Git repository that every bot treats as the source of truth. At ApexLogics I designed and rolled out AI-assisted development workflows and built automated pipelines that turn a written spec into a working build. Before AI tools, I spent 15 years building and shipping software in game development, and I bring the same habits to automation: clear contracts between steps, checks you can test, and an audit trail.

What I built, and what happened

A multi-agent AI fleet that runs my life and company admin

I designed and run a fleet of 24 AI agents. 18 are work bots, each with one narrow job: a dispatcher that talks to me and hands out work, standing routines for bills, inbox and finance, a Gmail worker, a job-search planner and a separate applier, a records filer, a web builder, a research bot, and a bot that designs and health-checks the other bots. The other six form a "thinking council" that handles big decisions: it frames the question, makes the best case for each option, red-teams the plan and checks it against reality before I choose. All of them work from one private Git repository that acts as the single source of truth. Every bot reads a map file before it touches anything, writes through the GitHub API in batches, and checks the result by commit SHA. Passwords stay in a password manager; the repo only holds pointers.

A job-search pipeline with a human in charge

The planner bot finds roles, scores them against my written preferences, and drafts an application packet. The applier bot can only submit after I approve in plain English and the dispatcher issues a release token that matches the packet's commit SHA. Every form it fills is logged field by field, with the source of each answer, so nothing is invented.

Company operations on agents

OnTriggerJoy itself runs on these agents: a Slack bot takes requests, files records, and writes documents such as the sales team's portfolio pack, and a company inbox bot sweeps company mail. I also built and shipped ontriggerjoy.com with an AI Automations section and a working enquiry form, plus my own portfolio and resume sites.

Skills

Agent design

  • Dispatcher and worker agents with narrow remits
  • Human approval gates and release tokens
  • Single source of truth with a map file
  • Batched writes, verify-after-write by SHA
  • Standing routines (bills, inbox, finance)

AI tooling

  • LLM coding agents with rules and skills (Cursor)
  • MCP servers (Unity MCP, Gemini via AI Studio)
  • Generative APIs (Ludo.ai, Suno)
  • AI-judged quality checks with written rules
  • Local image generation (ComfyUI)

Integrations

  • GitHub Contents and Git Data APIs
  • Gmail, Google Calendar, Slack
  • Hostinger web hosting
  • Firebase Analytics
  • JSON manifests checked against schemas

Engineering

  • C# (expert), Unity (expert)
  • HTML, CSS, JavaScript
  • Git and pull-request flow
  • Editor and build automation tools
  • Training teams on new workflows

Experience

OnTriggerJoyFeb 2024 – Present
Founder · Punjab, Pakistan · Remote

Founder. Built and run the AI agent fleet, the company website and its AI Automations offer; shipped Symphorix to Steam Early Access.

ApexLogicsMar 2026 – Oct 2026
Game Producer · Rawalpindi
  • Led production for the studio's mobile game portfolio, managing the whole development team, sprint planning, QA and releases.
  • Built an automated AI pipeline that turns game ideas into playable prototypes.
  • Rolled out AI-assisted development across the team and trained the developers to use it.
  • Moved the studio to modern source control and review-based workflows.
  • Used player analytics to find drop-off points and drove the fixes that improved player progression and retention in live games.
  • Built core gameplay systems and major performance improvements for live titles, including faster load times and crash fixes.
  • Wrote 20+ custom shaders across six games.
  • Built reusable tools and frameworks that the studio adopted across multiple games.
Monarch / GenopetsApr 2024 – Sep 2024
Lead Technical Artist & Procedural Systems Developer · Remote · Genopets (flagship)

Led procedural generation of the pets and terrains and wrote custom shaders that kept the game fast on phones.

TintashMar 2020 – Mar 2024
Game Producer · Technical Domain Expert — Games · Lahore · Remote

Producer and technical lead on several Unity client titles; lead PM and developer on Gamerize Dictionary.

OZI TechnologyOct 2019 – Mar 2020
Principal Software Engineer → Technology Lead · Lahore
UltPult Game StudioJan 2017 – Mar 2017
Head of Game Development · Lahore

Built an in-house Unity SDK that cut the studio's development time per game from over a year to under three months.

Giant Fox StudiosFeb 2016 – Dec 2016
Lead Game Developer · Lahore / Remote

Led Unity development; ported There's Poop In My Soup (K Bros Games) to mobile.

Education

MBA, Human Resources Management (UET Taxila / APCOMS) · BBA (Hons) (NUML / APCOMS)