About

Learn more about me

Mike Polinowski

AI-Driven Infrastructure, Computer Vision & Cloud Operations

A M.Sc. in Atmospheric Chemistry & Plasma Physics taught me to build systems where every variable matters. Sixteen years in the IT industry taught me that the same rigor scales from a 40 MHz microcontroller to a 70-node bare-metal fleet.


  • Nationality: German
  • Website: mpolinowski.github.io
  • Email: mpolinowski@gmail.com
  • Location: Hong Kong / Shen Zhen
  • Company: INSTAR Deutschland GmbH
  • Degree: M.Sc. Chemistry (Atmospheric Chemistry, Plasma Physics, Biological Chemistry)
  • Position: CTO
  • Email: m.polinowski@instar.com

The Trajectory

Domain Description
CTO, INSTAR Deutschland GmbH Owned DevOps, DevSecOps and platform strategy for 70+ bare-metal cloud servers and VPS instances serving both internal tooling and customer-facing web services.
AI / MLOps (last 5 yrs) Designed and deployed CVE (Computer Vision for Embedded) models on IoT edge hardware *and* customer-facing cloud APIs. Built ML pipelines from data ingestion through model serving with full observability.
IoT / Embedded QA → Full-Stack Web Dev → Network Automation 16 yrs of building, shipping, and running the full stack.
Most recent project End-to-end, CPU-constrained AI inference chain: YOLO-class object detection → face & ANPR (license-plate) recognition → a fine-tuned Vision-Language Model (VLM) for contextual scene description in security workflows. All running on resource-limited edge hardware.

What I bring to a new Role

Domain Description
DevOps / DevSecOps / MLOps CI/CD, IaC, service mesh, monitoring & alerting, security hardening – from the developer's laptop to production, cloud-native and observable.
Cloud & Linux at scale Bare-metal, VPS, Kubernetes-adjacent stacks, provisioning tooling, multi-tenant web services.
AI / CV / Edge ML Model selection, fine-tuning, quantisation for CPU-only targets, real-time inference on embedded and cloud tiers.
Leadership CTO-level ownership: strategy, vendor selection, team direction, cross-functional delivery.

Years of Computer Vision

Years of DevOps

Years Web Development

OSS Repositories

Tools I Like

Linux 100%
Docker 100%
Hashicorp Nomad 90%
Zabbix 90%
Python 90%
PyTorch 65%
Tensorflow 60%
Elasticsearch 85%
Kubernetes 55%
Typescript 80%
Node-RED 100%
Photoshop/Gimp 100%

Interests

Science

Linux

Photography

Travel

Frontend Design

Smart Automation

Connecting Services

Computer Vision

Machine Learning

Environmental Science

Open Source

IoT Connect

Resume

Check My Resume

Profile

Mike Polinowski

  • No.12069, Shennan Rd
  • Nanshan District, Shenzhen
  • m.polinowski@instar.com

M.Sc. in Chemistry & Plasma Physics with 16 years of progressive IT experience, spanning embedded IoT QA, full-stack web development, and cloud infrastructure at CTO level.

Last 5 years dedicated to AI / MLOps: building and deploying computer-vision and VLM inference chains on CPU-constrained edge hardware and cloud, with full CI/CD, observability, and DevSecOps ownership across a 70+ node bare-metal fleet (HashiCorp Nomad / Consul, Zabbix, self-hosted GitLab).

Strong at the intersection of systems thinking from the sciences, hands-on Linux/cloud operations, and applied ML – from a 40 MHz sensor node to a multi-tenant cloud platform.



Domain Tools & Technologies
AI / ML / CV PyTorch, TensorFlow, YOLO / EfficientDet, ANPR, VLMs (fine-tuned, quantised), transfer learning, ONNX Runtime, OpenVINO, edge inference
MLOps / DevOps GitLab CI/CD, HashiCorp Nomad & Consul, Docker / containers, Zabbix, Grafana, Prometheus, Terraform, IaC
Cloud / Infra Bare-metal Linux clusters (70+), VPS, CentOS / RHEL / Ubuntu, Nginx, HAProxy, multi-tenant web services
Web / Backend Node.js, React, Elasticsearch, REST / gRPC APIs, service-mesh patterns, search-engine backends
IoT / Embedded MQTT, ONVIF, REST Webhooks, ARM / MIPS targets, Open-Protocol integration, QA / burn-in on production line
Security DevSecOps hardening, CVE patching, security monitoring, access control on customer-facing services
Languages Python, JavaScript / TypeScript, Bash, Rust (embedded), Go, SQL

Education

Research Assistant & Applied Astrophysics / Plasma Physics

2005 - 2010

University of Cologne, Germany

Mass spectrometry & Terahertz spectroscopy for iron-compound detection in low-pressure, high-energy plasmas. Teaching position – inorganic-chemistry lab courses.

M.Sc. & Applied Astrophysics / Plasma Physics

2002 - 2004

University of Cologne, Germany

Mass spectrometry & Terahertz spectroscopy for iron-compound detection in low-pressure, high-energy plasmas. Teaching position – inorganic-chemistry lab courses.

B.Sc. - Atmospheric Science & Biological Chemistry

2000 - 2002

University of Wuppertal, Germany

Spectroscopic methods for atmospheric analysis (Physical Chemistry). Teaching position – instrumental analysis & inorganic chemistry.

Professional Experience

Chief Technology Officer

2016 – Present

INSTAR Deutschland GmbH (Waletech)

Hongkong S.A.R. / Shenzhen, China

Own the full technology stack – from embedded firmware QA through cloud infrastructure to production AI/ML services – for a video-surveillance and smart-automation manufacturer.

  • AI / MLOps
    • Architected an end-to-end, CPU-constrained AI inference chain running on edge hardware: real-time object detection → face & ANPR (license-plate) recognition → a fine-tuned Vision-Language Model (VLM) for contextual scene description in security workflows. Quantised for deployment without a GPU.
    • Built MLOps pipelines (data ingestion → training / fine-tuning → model registry → serving) with automated transfer-learning loops, version-controlled via GitLab CI/CD.
    • Designed and shipped computer-vision prediction APIs and search-engine backends (Elasticsearch) as customer-facing cloud services.
  • DevOps / DevSecOps / Cloud Infrastructure
    • Operate a 70+ bare-metal cloud-server and VPS fleet (Linux, CentOS / Ubuntu) serving internal tooling (online documentation, support-ticket platform, SysAdmin dashboards, marketing systems) and customer-facing web services.
    • Migrated the platform to HashiCorp Nomad + Consul for containerised and legacy workloads – self-healing, auto-scaling, delocalised across data centres.
    • Standardised CI/CD on self-hosted GitLab (containerisation, automated testing, blue-green deploys). Integrated Zabbix + Grafana / Prometheus for server- and application-layer observability.
    • Owned DevSecOps: patch management, CVE response, security monitoring, and access-control hardening across all customer-facing services.
  • IoT & Product Expansion
    • Extended INSTAR's product scope into home automation by integrating open IoT standards and communication protocols (MQTT, ONVIF, REST Webhooks) into the firmware and cloud stack.
    • Drove cross-functional delivery: coordinated 7-person teams, vendor selection, and architecture reviews as CTO.

General Manager

2012 - 2015

INSTAR Deutschland GmbH

Guangzhou, China

Managed a 5+ person team and 5 concurrent projects under tight deadlines while the company scaled its online presence and customer-support tooling.

  • Full Stack WebDev / Product QA
    • Led development and maintenance of a customer-facing knowledge-base platform (Node.js, React.js, CentOS, Elasticsearch backend); integrated the platform into all in-house software products.
    • Designed and rolled out internal training programmes for the customer-service and production teams.
    • Coordinated new-product development and QA in close cooperation with the production-line team; acted as liaison between engineering, manufacturing, and sales.
    • Built and maintained provisioning tooling and internal dashboards to streamline onboarding of new SKUs and regional deployments.

Support / QA Technician

2011 - 2012

INSTAR Deutschland GmbH

Guangzhou, China

First technical hire on the China side; covered the full support-and-QA cycle from bench to production line.

  • Full Stack WebDev / Product QA
    • Delivered Tier-1 / Tier-2 technical support to end customers (video-surveillance cameras, NVRs, network switches) and performed on-site quality assurance across the production cycle.
    • Authored customer-facing print and online documentation; created training material for new hires.
    • Enabled system integration of INSTAR hardware into third-party software and security platforms.
    • Performed firmware QA / burn-in testing on IoT and network devices; documented failure modes and fed findings back to the hardware design team.

Services

My Services

AI & Computer Vision

Real-time object detection, face & ANPR recognition, and fine-tuned Vision-Language Models — quantised and deployed on CPU-constrained edge hardware, no GPU required.

MLOps & Model Pipelines

End-to-end ML lifecycle: automated transfer learning, model registry, version-controlled training runs, and production inference serving — all driven through CI/CD.

Cloud Infrastructure

Self-healing, auto-scaling bare-metal clusters (70+ nodes) orchestrated with HashiCorp Nomad & Consul. Multi-tenant web services, internal tooling, and customer-facing APIs.

DevOps & DevSecOps

GitLab CI/CD, containerised deployments, and security hardening: CVE response, security monitoring, and access-control across every customer-facing service.

IoT & Embedded Systems

Open-protocol integration (MQTT, ONVIF, Modbus) for surveillance and smart-automation hardware. Firmware QA on ARM/MIPS targets and production-line validation.

Web Platforms & Observability

Node.js / React frontends, Elasticsearch backends, and REST APIs — instrumented end-to-end with Zabbix, Grafana, and Prometheus.

Portfolio

My Projects

  • All
  • MLOps
  • DevOps
  • SysOps
  • WebDev
  • IoT
INSTAR Cloud Computer Vision

CPU-Only AI Inference Pipeline

Multi-GPU fleet → one CPU server, 10 K+ det/min

INSTAR Cloud CVE Pipeline

INSTAR Cloud AI Pipeline

Full CV chain, stable, CPU-only

INSTAR Cloud Vision Language Annotation

Edge Vision-Language Model

Fine-tuned VLM on CPU, OpenVINO-exported

INSTAR Cloud License Plate Recognition

ANPR – Fast-Plate-OCR

Single-digit-ms plate reads, no GPU

INSTAR Cloud Face Recognition

Face Recognition Pipeline

Customer-triggered training via the Cloud UI

Hashicorp Nomad

Unified Deployment Orchestration

70+ nodes, self-healing, one-PR onboarding

Zabbix Server Monitor

Fleet Observability & Alerting

Know it's breaking before the customer does

Elastic Stack | Enterprise Search

Search & Analytics Backend

Docs, AI events, tickets — one index

INSTAR Wiki | Knowledgebase

Customer Knowledge Base

Zero docs → self-service customer search

Discourse | Customer Forum

Community Forum

Rep tickets → community answers themselves

osTicket | Support Ticketing System

Support Ticketing System

Inbox chaos → SLA-tracked support queue

MQTT Control API - The Universal Gateway to Automation

Smart-Home Protocol Integration

Only camera maker with full MQTT control

NTS | Secure Timeserver

Fleet Time Synchronisation

One clock for 70+ cameras and servers

Gitlab | The DevSecOps Platform

CI/CD & Source Control

Merge → container → production, automated

Node-RED | Programming for event-driven applications

Ecosystem QA & Tier-2 Support

Still the go-to escalation, 5+ years in

Contact

Contact Me

My Address

Rm A1912, Coastal Times Plaza

No.12069, Shennan Rd

Nanshan District, Shenzhen

Social Profiles

Email Me

mpolinowski@gmail.com

Github

Profile

Designed with BootstrapMade