Recoder
AI-assisted game regeneration systemGrounds agents in real project evidence, coordinates engineering tools, checks runtime behavior, and produces testable builds.
Result: 4–5 generated games, 2 released to stores.
Hai Tran Ngoc · Applied AI Engineer
It automates most of the reconstruction workflow: grounding AI agents in real project evidence, rebuilding gameplay in dependency order, and checking candidate behavior against runtime traces.
Selected work
Grounds agents in real project evidence, coordinates engineering tools, checks runtime behavior, and produces testable builds.
Result: 4–5 generated games, 2 released to stores.
Video file will be added here
Company experience
The work is described by problem, contribution, and outcome—without exposing company data or proprietary implementation details.
Built a unified workflow that reads creative assets from shared storage, applies naming and mapping rules, and prepares bulk delivery to Google Ads and Unity Ads. AI is used only for tasks that need judgment—such as generating ad copy or identifying a creative type—while deterministic steps stay rule-based for speed and reliability.
Contribution: workflow design, platform integrations, batch processing, and review-ready outputs.
Automated the path from localized metadata to draft submission on Google Play and App Store Connect. The workflow handles store-specific constraints, including API permissions, metadata and asset upload, and Apple version creation before localization.
Contribution: API integration, AI-assisted metadata generation, versioning rules, and safe human review gates.
Built tools to inspect and modify playable ads, replace assets and store links, edit board data, and verify the final render. I also designed level-generation workflows for tile- and arrow-based puzzle games so designers could create and test content faster.
Contribution: tooling architecture, runtime hooks, AI-facing tool interfaces, and automated verification.
Designed AI-assisted workflows across recruitment, onboarding, internal knowledge, contracts, attendance, payroll, reminders, and offboarding. The system follows deadlines, drafts routine content, detects missing or unusual cases, and escalates work while keeping final decisions with HR.
Contribution: workflow modelling, agent boundaries, SLA and escalation rules, and human-in-the-loop design.
Contributed to production pipelines that bring RevenueCat and Adjust data into BigQuery for reporting and analysis. I also worked on large-scale crawling flows, data normalization, scheduled processing, and quality checks so downstream teams could rely on consistent datasets.
Contribution: ingestion pipelines, orchestration, data validation, and operational monitoring.
Core stack
Agents that reason, build, and verify game behavior.
APIs, services, and data systems that keep workflows reliable.
Containerized deployments and VM operations across GCP and AWS.
AI coding and reasoning tools used across research, implementation, and review.
Static and dynamic analysis across Android, native binaries, Unity, and IL2CPP runtimes.
Background
Building AI-assisted systems across game production, internal operations, data engineering, and cloud delivery.
Gameplay systems, custom tools, UI, debugging, and optimization.
Frontend development, API integration, Unity instruction, and mentoring.