Skip to content
View ioanlucut's full-sized avatar

Highlights

  • Pro

Block or report ioanlucut

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ioanlucut/README.md

Hi, I’m Ioan 👋

I’m a Lead Engineer at ImmoScout24 Austria, building search platforms and APIs. I’ve been with the company for 10+ years.

I’m a dad of four (yes, four), into music and guitar, and a Home Assistant enthusiast. After eight years in Austria, I’m now based in Sibiu, Romania.

I started in control theory and automation engineering, then found my place in software. AI workflows have brought back my love of automation. Apparently, I’ve come full circle. Or closed the loop.

At work I still build things. With my fellow leads and our Head of Technology, I also help decide where our Austrian engineering goes: architecture, engineering standards and what we deliver first.

Right now I’m focused on AI-native engineering: agents that get the context they need, workflows we can reuse, and fast verification. People stay in control.

Things I’ve built and helped shape

  • Property-search platform — I led the architecture of the 2018–2019 relaunch and built core parts of it: streaming React SSR, the first GraphQL search API and the Elasticsearch-backed search URLs. Then I led the phased move of the legacy Angular search and Java catalogue onto it. I’ve kept evolving it since, including a rebuilt geographic autocomplete and shape-based regional search. It’s still in production.

  • Map search — The map shows a sample of real properties at their actual locations, not cluster centres. I led delivery from feasibility to first production release in three months (2025) and built the Elasticsearch aggregation behind it; its p95 stayed under 100 ms in production (team metrics, August–November 2025). Later, customers reported hidden-address listings showing at their true location, so I designed and built stable masking: the pin stays inside the right postal-code area and doesn’t move between visits.

  • AI-assisted fraud review — Customer Service used to check private listings by hand before publication, weekends included. I led the Dify/Claude on Bedrock integration with them and another engineer: they brought the fraud expertise and the prompt, and manual fallback stays in place. In its June–August 2026 report, Support had 50% of those checks clearing for immediate publication.

  • AI recommendations — Live in the Austrian apps since September 2026. I led the Austrian side: data contracts, identity stitching, event pipelines and the app-feed rollout. The German AI Solutions team owned the embedding model, SageMaker API and OpenSearch vector search.

  • Natural-language search — The Austrian team’s query tokenizer turns free-text German property searches into structured search parameters. My part: early spaCy exploration, architecture and design feedback, and helping bring the service into production.

  • Recoverable checkouts — A failed Salesforce integration step used to mean manual recovery. I rebuilt checkout completion as visible, retryable workflows on Step Functions and Lambda. From July 2025 through September 2026, CloudWatch recorded 50,000+ successful workflow completions across paid and free checkouts, with 4 failures and 2 timeouts.

  • Shared frontend foundations — I co-created and still maintain the shared React components and build tooling that 23 Austrian applications used in 2026.

  • Agent tooling — I built versioned rules and reusable review and migration skills for coding agents, plus a tested tool that rewrites legacy Storybook component examples into the current story format. We used it in the Storybook 6→10 upgrade.

Open source & writing

Side projects, old and new

  • Cortex — My Home Assistant setup as code, including solar-surplus EV charging and home-energy controls. I use AI agents to audit the setup and develop changes through tests, review and automated deployment.

  • bes-lyrics — Lyrics and chords as code. I built the checks and songbook tooling; others maintain the content.

  • Remo — Started as my 2014 master’s dissertation: distributed process control through web services.

  • Revaluate · 2015–2017 — Expense tracking: my Java API and most AngularJS logic, a collaborator’s design and most styling.

Usually somewhere around: TypeScript, React, Node.js, GraphQL, Elasticsearch and AWS—with Claude Code, Codex and Datadog in the mix.

CV (PDF) · LinkedIn · ioan.lucut88@gmail.com

Pinned Loading

  1. bes-propres7-migrator bes-propres7-migrator Public

    Songs as code for ProPresenter 7: a Git-managed song library compiled into native .pro presentations via the reverse-engineered protobuf format, shipping only what changed to the presentation Mac.

    TypeScript 5

  2. revaluate-web revaluate-web Public

    Personal finance web app for logging expenses, spending insights and monthly goals. AngularJS SPA, launched 2015 (archived).

    JavaScript

  3. reme reme Public

    Email reminders you write as a sentence. A side project I co-founded in 2014, built after hours, and brought back to life in 2026.

    JavaScript 1

  4. revaluate-api revaluate-api Public

    Backend of Revaluate, a personal expense tracker launched in 2015: expenses, goals, insights, Mint and Spendee imports, a Slack command and Braintree subscriptions. Dropwizard, Spring, JPA and Post…

    Java 1

  5. bes-lyrics bes-lyrics Public

    Songs as code: 1,935 Romanian worship songs as plain text in Git, reviewed in pull requests, statically checked by CI, and compiled to ProPresenter 7 and a PDF songbook with chords.

    TypeScript 6

  6. remo remo Public

    PID control across an unreliable network: a remote sensor and actuator with fault injection, fail-safe recovery and a live dashboard. Master's thesis (TU Cluj-Napoca, 2014), rebuilt in Java 21.

    Java