K hetzenauer.me
Available for work — Software Developer

Hi, I'm

Kevin Hetzenauer

Software developer building enterprise software at scale — from AI workflows and integrations to complex domains with millions of records, engineered to stay safe and auditable.

View work Get in touch
6+
years experience
1.3M+
records at scale
95%
faster search
<1s
query time
Full-stack
end-to-end
Angular TypeScript OpenSearch MySQL Debezium CDC Machine Learning Vector Search Gemini API LLM Integration React AWS Next.js Rust Federated Learning Python Java Vaadin
01

What I do

I build enterprise software at scale — from AI workflows and integrations to complex domains with millions of records and real-time data. I care about data modelling done right, state machines you can audit, and concurrency-safe design that holds up under real load.

Frontend
Product engineering
Angular 20TypeScriptSignalsReactNext.jsRxJS
backend & data
Data & backend
NestJSMySQLKyselyOpenSearchDebezium / CDCEvent-driven
architecture
Systems design
State machinesConcurrency & lockingReal-time / WebSocketsDomain modellingIdempotencyAudit trails
ai & ml
AI & research
LLM IntegrationGemini APIMachine LearningFederated LearningEdge ComputingPython
02

Things I've built

A cross-section of my work — enterprise systems, data, AI features and side projects. It's the range, not any single project, that shows the kind of developer I am. Click any card for details.

01 — Systems Architecture

In a manufacturing ERP a part is never just deleted — it can be active in one country, obsolete in another, sitting in valuable stock, and wired into thousands of live bills of materials. The legacy system flattened all of that into a single global status flag shared by five companies. I replaced it with a per-client status model and built a guided, multi-phase workflow that retires a part and migrates its BOMs to a successor — per client, with a complete decision history — safe to run on a live, million-part catalogue.

1.07M
item records governed
13.5M
BOM positions in scope
643K
per-client status rows
5
companies, one catalogue
7+4
phases + outcomes
Angular 20TypeScriptNestJSMySQLKyselyState MachineWebSocketsEvent-Driven
02 — Enterprise Software

Developed a high-performance item database solution at 3CON (3Connect ERP Software), handling over 1.3 million inventory items. Implemented seamless integration between MySQL and OpenSearch to enable sub-second search response times. The system features real-time data synchronization using Debezium's CDC technology to maintain search index accuracy. Achieved search speeds of <1s (simple queries) to 4s (complex joins), down from 20–30s previously — improving efficiency by over 95%.

AngularTypeScriptOpenSearchMySQLDebeziumCDCEnterprise Software
03 — Data Quality

Built a similarity-search and deduplication feature on top of OpenSearch, combining vector search and machine learning to surface near-duplicate items across a catalogue of over a million records — significantly improving data quality and reducing redundant entries.

OpenSearchMachine LearningTypeScriptVector SearchData Quality
04 — AI Feature

An LLM-powered feature that generates and standardizes item names and translates them across languages, integrating the Gemini API directly into the ERP's Angular interface to keep catalogue naming consistent and multilingual.

AngularTypeScriptMySQLLLM IntegrationAI-Powered TranslationGemini API
05 — Process Optimization

Designed and built a system to manage the full lifecycle of technical drawings — from creation through review and approval — streamlining workflows and reducing manual process overhead across teams.

AngularProcess OptimizationLifecycle Management
06 — Mobile App

A paragliding tracking and booking app built with React and AWS, covering flight logging and tariff booking for pilots and schools. Now discontinued.

ReactAWS
03

Studies & research

Academic and side projects that shaped my skills — from federated learning research to apps I shipped and shelved.

S1 — B.Sc. Thesis

An experimental investigation into the feasibility of Federated Learning on Internet-of-Things devices, built around a proof of concept comparing centralized and decentralized approaches. Completed as part of the B.Sc. in Digital Business & Software Engineering at MCI.

Federated LearningIoT SecurityMQTTEdge ComputingPrivacy-Preserving MLPythonRaspberry Pi
S2 — Side Project

A progressive web app for food and inventory management with usage analytics, built with Next.js and a Rust backend. Now discontinued.

PWANext JSRustInventory ManagementAnalytics
S3 — Side Project

An attendance-management app with GPS-based check-in verification, built with Vaadin and Java. Now discontinued.

VaadinAttendance SystemGPS VerificationJava