Carl Tseng

01 Impact

• ~90%compute cost cut on K8s test platform
• 30×test concurrency
• 17h → 50mfull-suite execution time
• $36K/yrCloudWatch spend eliminated
• 827K+APM data drops eliminated
• 400 → 20mscross-border streaming latency

02 Experience

1. Sep 2019 – Present

   Site Reliability Engineer @ Digital Forest Technologies

   💰 Cost & Infrastructure Optimization

   • K8s Architecture & Compute Optimization~90% (saving ~$3,300 annually)30x95% (from 17 hours to under 50 minutes)

     Engineered a transition from a monolithic EKS deployment to an on-demand "Job-per-game" testing architecture for intensive browser workloads. Mitigated severe K8s node CPU throttling (70% scheduling overhead) and eliminated a 21% timeout rate by deploying dedicated Karpenter NodePools with static CPU policies. Leveraged scale-to-zero automation to slash compute costs by ~90% (saving ~$3,300 annually), while boosting test concurrency by 30x and reducing full-suite execution time by 95% (from 17 hours to under 50 minutes).

   • AI & Storage Cost Reduction95% reduction

     Achieved 95% reduction in both AI inference and daily storage costs through LLM architectural redesigns and tiered storage optimization.

   • Observability Cost Savings$36,000 USD annually

     Saved $36,000 USD annually on AWS CloudWatch by replacing redundant Elasticsearch polling with on-demand Grafana plugins, maintaining full visibility for 10 active panels.

   ☁️ Cloud Architecture & CI/CD

   • Cloud Infrastructure & Migration

     Architected production AWS environments (EKS, EC2, S3) and executed seamless on-premise to AWS migrations with zero monitoring gaps.

   • Pipeline & Automation Engineering

     Engineered automated Azure DevOps pipelines for Docker/ECR publishing and zero-downtime EKS deployments, alongside AWX/Ansible automation for infrastructure lifecycle management.

   • Automated Telemetry Provisioning

     Engineered a tag-driven deployment mechanism via AWS SSM and Azure DevOps to migrate EC2 fleets from Elastic to OpenTelemetry agents across 6 AWS accounts. Achieved zero-touch installations and dynamic configuration management.

   📊 Observability & Data Integrity

   • External Partner Monitoring

     Built an end-to-end ELK pipeline (Filebeat/Logstash/Kibana) to monitor external studio services. Automated log parsing and anomaly alerting, enabling tier-1 teams to provide proactive technical support and ensure partner SLA compliance.

   • Telemetry Pipeline Resiliency827K+ drops

     Eliminated massive APM data loss (827K+ drops) by redesigning OpenTelemetry pipelines with message queuing to buffer ELK I/O bottlenecks. Fortified node stability using K8s taints, tolerations, and topology limits.

   • Time-Series Data Integrity

     Resolved data gaps in Grafana caused by SQL execution lags (>60s). Architected a stateful reference table to track timestamps, ensuring sequential, gap-less data backfills and 100% chronological accuracy.

   • Automated Fleet Provisioning & Monitoring

     Orchestrated a zero-touch OpenTelemetry agent deployment framework using AWS SSM and tag-based associations. Identified and resolved critical gaps in Auto Scaling Group (ASG) lifecycle events and IMDS tag propagation by collaborating with Terraform engineers, guaranteeing seamless, automated monitoring for all dynamically scaled infrastructure.

   🚀 Performance Tuning

   • Multi-CDN & Latency Optimization400+ms to ~20ms90%

     Resolved cross-border streaming outages by integrating Akamai/CDNetworks. Slashed latency from 400+ms to ~20ms, reduced packet loss by 90%, and optimized TTFB to meet strict SLOs (Estonia/Bulgaria/Colombia).

   • High Availability Streaming Architecture

     Engineered an automated cross-region failover mechanism for live video streaming. Designed a resilient routing strategy from Estonia service providers to AWS Frankfurt (primary), with automatic fallback to AWS Stockholm (secondary), guaranteeing zero-downtime service stability.

   • Automated Failover & Database Tuning

     Engineered a dynamic CDN failover system triggered by 10-minute rolling health checks. Concurrently profiled SQL stored procedures to eliminate processing delays and implemented primary/secondary HA failovers for zero-interruption service.

   🌏 Reliability & Global Resilience

   • Automated CDN Routing & Failover

     Prevented outages by developing a script to monitor multi-CDN network health. Streamlined dynamic DNS switching by syncing optimal routing states to AWS S3, enabling automated, real-time backend node failovers.

   • High Availability & Disaster Recovery

     Maintained production SLA by designing robust backup strategies, rapid incident response playbooks, and automated failover mechanisms to ensure business continuity.

   • Global Operations Leadership

     Acted as sole overseas technical lead (Sep 2021–Apr 2022). Mentored local hires, maintained cross-border IT stability, and delivered a 3-month engineering onboarding program entirely in English.

   stack: Azure OpenAI, Playwright, AWS (EKS, EC2, Lambda, API Gateway, S3), VMC, ELK Stack, Grafana, Ansible/AWX, Octopus Deploy, Git, PowerShell, Bash, Windows/Linux, IIS, MSSQL, Python, PHP, Java

2. May 2017 – Apr 2019

   Product Engineer @ BIOSTAR MICRO-Technology

   ⚙️ Global Operations & Hardware Automation

   • Global Fleet Management2,500+ mining rigs

     Led the development of a centralized management system to monitor and operate 2,500+ mining rigs across Japan, China, and Canada, significantly improving global operational uptime and stability.

   • Automation & Hardware Optimization

     Streamlined deployment and maintenance workflows by authoring custom Batch scripts for zero-touch system setup and real-time thermal monitoring. Fine-tuned GPU cooling parameters to maximize hardware longevity and optimize the power-to-performance ratio.

   • End-to-End Client Solutions

     Delivered tailored mining infrastructure solutions for Japanese clients, managing the complete lifecycle from hardware specification and technical definition to final configuration.

   • Cross-Functional & Regional Leadership

     Partnered with R&D and Sales teams to refine product specifications based on market feedback and cost-profit analyses. Provided expert technical oversight for daily operations in China, ensuring strict adherence to performance targets.

3. Aug 2015 – Apr 2017

   System Engineer @ EBM Technologies

   🏥 Clinical Infrastructure

   • Clinical Infrastructure & Database Operations

     Managed high-availability medical imaging (PACS) infrastructures across Windows Server and MSSQL environments. Orchestrated seamless database migrations, version upgrades, and robust disaster recovery procedures via automated snapshots and offsite backups.

   • Cross-System Integration (POC Leadership)

     Spearheaded a strategic PACS Proof-of-Concept, successfully integrating medical device telemetry and live video streams into the Hospital Information System (HIS) to establish an end-to-end clinical data workflow.

   • Incident Response & Hospital IT Reliability

     Delivered dedicated on-site hospital support to guarantee PACS stability and cross-departmental data integrity. Reduced "Time to Clarity" during major outages by translating technical system metrics into service availability reports, prioritizing triage based strictly on clinical and patient impact.

03 Side Projects

• 天堂 ETF — Lineage of Investment

  ↗

  lineage-etf.site

  A gamified, RPG-style planner for Taiwan ETF investing. Model loan-funded dollar-cost averaging, track holdings against live TWSE prices, and forecast break-even and real-profit dates after interest.

  • TWSE OpenAPI
  • Gamification
  • Cloudflare

• Home Network Monitor

  Pings 4 targets every minute and runs a full speed test only when latency or packet loss turns abnormal, plus one on the hour. Cut test traffic from 150–300 GB/day to about 12–36 GB/day, keeps 30 days in Prometheus/Grafana, and generates an evidence report for the ISP.

  • Docker Compose
  • Prometheus
  • Grafana
  • Python

• Sleep & Health Analytics

  ↗

  github.com/CarlJune4/Healthy_Prediction

  Streams Apple Health exports straight from the zip (about 100× less memory than batch parsing) to relate previous-night sleep to next-day heart rate and workouts. A Streamlit app trains a personalized model from an uploaded export.

  • Python
  • pandas
  • scikit-learn
  • Streamlit

• Insta360 Competitor Price Tracker

  Graduate course project. n8n runs browser-use agents on a schedule to scrape competitor prices, cleans and normalizes the results, indexes them into Elasticsearch, and tracks price movements on Grafana dashboards.

  • Course project
  • n8n
  • browser-use
  • Elasticsearch
  • Grafana

04 Skills

05 Education & Certifications

06 Off-screen

• 🏋️HYROX2× finisher

  8 km of running split by 8 functional stations. Pacing, recovery, and finishing under fatigue: the same habits that hold up through a long incident.

• 🤿Scuba diving120+ dives

  Certified scuba diver. Plan the dive, check the gear twice, and stay calm when something goes wrong.