Schwannden Kuo
Tech Lead, Identity & Applied AI
Professional Summary
Tech lead with a decade across data products, identity platforms, and applied AI. Currently leading SSO subsystem at Ubiquiti serving millions of active users under high-throughput production load, and built the team's portfolio of production AI agents — multi-agent triage, release monitoring, code review — that learn from each use. Previously Director of ML at MoBagel, where I won the enterprise accounts (Chunghwa Telecom, Wistron, AUO) that anchored Series A revenue and re-architected the AutoML engine for ~30× runtime efficiency. Author, speaker, open-source contributor.
Work Experience
SSO Team Lead
Ubiquiti • Aug 2025 - Present
- • Built and shipped a portfolio of production AI agents for the SSO team — multi-agent release monitor, read-only incident-triage agent spanning CloudWatch, RDS Performance Insights, Athena, ALB/WAF, and CloudTrail, plus team-shared skills for code review, dependency review, and support triage. Triage agent cut median alert-to-root-cause from 70 → 18 minutes across 5 production incidents; by the fifth, a teammate ran incidents solo. Documented in a 3-part public blog series.
- • Drove a multi-quarter API latency epic to completion — identified SES suppression-list lookups as the root cause and shipped a Redis-cached fail-safe. P95 reduced 90–97% across 6 critical auth/MFA endpoints (e.g., login 11.27s → 0.79s; email MFA 10.18s → 0.31s; password reset 9.27s → 0.54s).
- • Owned cross-team architectural delivery on the SSO platform — drove Bring-Your-Own-IdP (SAML/OIDC for Entra, Okta, Google Workspace) from zero prior protocol experience, with the Entra one-click flow now live; brokered cross-team contracts across SRE, IdP Cloud, MSP Mgmt, Account FE, on-premise devices, and Mobile FE for multi-site tenant provisioning, JWT key rotation, and unified API error response standardization.
Cloud Software Engineer
Ubiquiti • Aug 2024 - Aug 2025
- • Velocity: reduced feature lead time ~30% and accelerated security patching ~5× by revamping the release lifecycle and automating consistency checks.
- • Resilience: diagnosed a 30-minute production login p95 spike (5.19s vs 1.5s baseline) as a Postgres row lock held across PBKDF2/Argon2 hashing; 15-LOC fix achieved a 697× improvement in mean exec time on hot-row updates.
- • Resilience (DB): engineered high-availability database architecture with strict read/write separation, eliminating production impact during RDS replica events.
- • Operated AWS SSO at millions of active users under high-throughput production load, aligning security architecture with business outcomes and reducing support cost.
Senior Software Engineer
Dell • Jan 2023 - Aug 2024
- • Modernized a legacy integration test platform into a cloud-native architecture (AWS & Kubernetes), reducing operational costs by 50% and accelerating product build time by 400%, directly improving time-to-market.
- • Advocated for developer-first practices by gathering feedback from engineering teams and influencing CI/CD adoption (GitHub Actions, Argo, Kustomize, Helm).
- • Established cloud security best practices (automated vulnerability scanning, SBOM, Trivy) to align engineering efficiency with enterprise compliance.
Senior Software Architect
MoBagel • Sep 2021 - Dec 2022
- • Architected 8ndpoint.com, the AI-driven advertising platform whose technical assets and customer traction anchored MoBagel's strategic M&A acquisition; integrated generative personalization for campaign engagement.
- • Led the migration of core services to a hybrid-cloud environment, designing scalable IAM and network structures across GCP (GKE, Cloud SQL, BigQuery) and AWS (EC2, S3, VPC) utilizing orchestration tools like ArgoCD and Airflow.
- • Partnered with product leadership to translate complex AI/ML capabilities into customer-facing benefits, effectively bridging the gap between R&D innovation and enterprise campaign ROI.
Director of Machine Learning
MoBagel • Jun 2018 - Sep 2021
- • Won three Taiwan enterprise accounts that anchored Series A revenue thesis — Chunghwa Telecom, Wistron, and AUO; two became strategic investors. Each engagement combined direct customer ownership with the technical solution: re-architected the general AutoML engine (Scala / Hadoop) for memory and compute efficiency to ship to Chunghwa Telecom — ~30× runtime improvement; designed bespoke algorithmic solutions for Wistron and AUO industrial problems with measured ROI per engagement.
- • Built the engineering organization from the ground up as hiring manager — interviewed and onboarded engineers across QA, DevOps, Web, and Algorithm functions; coached the Frontend Lead and Algorithm Team Lead through their first 12 months; transitioned the QA function to a hired QA Lead.
- • Restructured the product to enable an "inner-source" culture across teams, accelerating feature rollout ~4×.
Senior Data Scientist
MoBagel • Jun 2017 - Jun 2018
- • Automated machine learning design and reactive machine learning architecture with Scala, Akka, Spark.
- • Design and implement the first generation AutoML system, achieve 300x speed up to secure key accounts.
Software Developer
University of Pittsburgh • Jun 2016 - Sep 2016
- • Integrate latest statistical methods in -omic study to build user friendly application for biostatisticians. Create open source project, to encourage open source collaboration and accelerate impact. Train researchers to maintain the project.
- • Publish paper on biostatistics (See publications).
Research Assistant
Academia Sinica • Jul 2014 - Jul 2015
- • Assist research in NDN network. Open sourced project (source code and book) to deploy NDN on Galileo. Featured in intel maker community.
Speaking Highlights
AI Agentic Development & Harness Engineering
Three-hour training across three parts. Part 1: paradigm shift to harness engineering — context engineering, skills as durable assets, the four eras of AI development. Part 2: bootstrapping the AI workflow — pre-commit/CI as agent sensory organs, skill authoring, plan-driven feature development. Part 3: provider-agnostic harness — portable artifacts (AGENTS.md, skills, MCP) across Claude, Codex, Cursor, and Gemini.
Audience: MoBagel core developer team and agentic AI team (~15 engineers)
End‑to‑End AutoML in Practice (Decanter AI)
Practical AutoML pipeline design, speed/accuracy trade‑offs, and real‑world deployments
Audience: Industry practitioners and partners; workshop attracted hundreds of participants
AIoT and Data Analysis: Approaching Industrial AI
Key principles of 'ML Done Right' including speed/accuracy trade-offs, evaluation vs. implementation cost, human vs. machine decision-making, and real-world case studies
Audience: Industry practitioners, partners, academics
Statistical Machine Learning Workshop (Campus‑wide)
Intuitive approaches to high‑dimensional statistics for engineering students; emphasized how statistical thinking underpins modern AI/ML
Audience: Students
And 3 more speaking engagements...
Technical Skills
Cloud & DevOps
Languages & Frameworks
Data & ML
Publications & Writings
Explore my technical publications, speaking engagements, and thought leadership contributions spanning cloud native technologies, AI/ML, and DevOps.
View Speaking & PublicationsOpen Source Contributions
View all contributions →Research Tools
Community Projects
Educational documentation site for building autonomous AI agents using Claude
+1 more features
And 1 more community projects...
Active contributor to various open source projects in the cloud native, Python, and DevOps ecosystems.
Education & Volunteer Experience
Education
Master's Degree
National Chiao Tung University • 2017
- • Teaching Assistant for queueing theory, machine learning, and formal language
- • Organizer and Speaker for 2017 NCTU Statistical Machine Learning Workshop
Bachelor's Degree
National Chiao Tung University • 2015
Volunteer Experience
Teacher for Special Kids (陪讀老師)
樂服社區關懷協會 - 兒少據點 • Aug 2017 - Feb 2019
Help kids on their homework, but most importantly just being there and let them know someone does care and love them.