AWS Certified Solutions Architect – Associate (SAA)
Amazon Web Services
Software Engineer, AI Innovation & Automation @ Roku|London, UK
Production AI Agents, Multi-Agent Systems and Autonomous Testing.
I’m Arkaan Quanunga, a software engineer at Roku building production AI agents across codebases, developer tools and physical devices. My work covers multi-agent orchestration, persistent state and memory, MCP tool interfaces, retrieval, isolated execution and validation. Explore QAntum, Explorer and Forge, alongside my earlier AWS and video infrastructure work at V-Nova.
~100 engineersQAntum adoption, from August 2026May 2025Explorer in production16,266Explorer validation signals99%PaPeR SPIE setup reduction, 3 hours to seconds
Ownership: ~4,000 commits and ~70% of Explorer version-control history, across a 17-engineer contributor base.
Credentials
Certifications
AWS Certified Machine Learning – Specialty (MLS)
Amazon Web Services
TensorFlow Developer Certificate
Google Developers Certification
ISTQB Certified Tester Foundation Level (CTFL)
ISTQB / BCS
IBM Data Science Professional Certificate
IBM
Sep 2020 – Jan 2021
Recognition
Patent pending and a published NVIDIA article
Patent pending · US & EU
Autonomous software verification
The Roku stack — Explorer, QAntum, and Forge — is a patent-pending invention for turning manual QA into a data-driven process that finds functional and visual defects on devices with less manual oversight.
Read it on the career timelineNVIDIA Developer Blog
Enabling Customizable GPU-Accelerated Video Transcoding Pipelines
Written with NVIDIA engineers at V-Nova. The article covers VC6 codec performance and a GPU-accelerated transcoding pipeline, published on NVIDIA’s developer blog.
Read the NVIDIA articleIntro
Intro
Core Engineering Deliverables
Featured Architecture Case Studies
Multi-Agent SystemsAug 2026 – Present
QANTUM
Slack-Native Multi-Agent QA Orchestrator
A Slack-native multi-agent QA orchestration platform routing engineering and QA requests across specialized agents via an extensible capability protocol. Features shared and user memory, Model Context Protocol (MCP) integrations, and autonomous MR/PR generation and validation flows—backed by production usage across ~100 engineers.
100Engineers
5,199Agent runs
0.17%Run failure rate
PythonClaude Agent SDKMCP ProtocolLiteLLM+4 more
Read QAntum’s multi-agent architectureAutonomous Systems & RLMay 2025 – Present (~16 months in production)
Roku Explorer
Autonomous Exploratory Testing Platform
A production autonomous QA platform for firmware and Grand Central product areas. Combines PPO/LSTM reinforcement learning, ChromaDB multimodal RAG, dual validators, and MCP-based device control to navigate device UIs, surface potential issues, and flag visual regressions with minimal human oversight.
Multi-serviceProduction stack
RL + RAGArchitecture
MCPDevice control
PythonPyTorch (PPO / LSTM)ChromaDBFlask+6 more
Explore autonomous testing with ExplorerAI Test Automation2026 – Present
Forge
AI Test-Authoring & Automation Agent
An intelligent AI automation agent that investigates product-area workstreams, creates detailed automation tickets, and autonomously generates, validates, and runs regression test suites across repositories using isolated git worktrees.
AutonomousWorkflow
WorktreesIsolation
Gated CIQuality
PythonGit WorktreesPytestLLM Reasoning+3 more
See Forge’s AI test-generation workflowDistributed Cloud InfrastructureNov 2022 – Dec 2023
PaPeR (Parallel Performance Runner)
Distributed Video Compression Infrastructure
High-throughput distributed cloud infrastructure built on AWS with Terraform, collaborating with Meta, Intel, and SPIE to automate video compression quality evaluation across thousands of parallel transcoding jobs.
-99%Setup Time
1,800+Hours Saved
+50%Throughput
AWS (EC2, S3, RDS, Lambda)TerraformPythonFlask+4 more
Read PaPeR’s distributed video benchmarkingCURRENT PRODUCTION WORK · 2025–PRESENT
Transforming Manual QA into an Autonomous, Data-Driven Pipeline
At Roku, I develop and deploy the next evolution of quality engineering: agent systems that autonomously navigate physical device interfaces, author robust regression tests, and orchestrate across complex developer tools. This unified stack connects code changes directly to risk-based automated physical device exploration.
Systems & Stack Matrix
Technical Capabilities & Engineering Domains
Proven across production multi-agent systems, large-scale video transcoding cloud infrastructure, and real-time computer vision hardware.
Autonomous QA & RL
PPO and LSTM reinforcement learning policies combined with multimodal vision models and ChromaDB vector retrieval to navigate state spaces autonomously.
PPO / LSTMChromaDBVision LLMs
Multi-Agent Systems
Capability routing protocols, Model Context Protocol (MCP) integrations, isolated agent memory, and automated code review Remediator loops.
MCP ProtocolClaude SDKSlack Bolt
Distributed Cloud & Video
Scalable AWS infrastructure automated with Terraform, FFmpeg pipelines, GPU transcoding acceleration (NVIDIA DevBlog), and VMAF perceptual quality benchmarking.
AWS TerraformDocker / K8sFFmpeg / CUDA
Edge Vision & AR
Real-time object detection (YOLO26), MediaPipe hand gesture tracking, monocle optical pipelines, and spatial glass HUD compositors.
YOLO26MediaPipeEmbedded Linux
Interested in Autonomous QA or Multi-Agent Architectures?
Contact me about AI engineering, agent infrastructure or technical collaborations.