Two decades of shipping software.

Reedy is a publishing industry app that brings backlist catalogs up to top-quality metadata. It analyzes market trends, generates SEO keywords and genre classifications, assigns standardized THEMA subject codes, and rewrites descriptions for discoverability. Every suggestion runs through a human-in-the-loop step where an editor accepts, edits or rejects it, and it integrates directly with ONIX files so publishers work against their existing catalog with no local install.

I worked across the platform, improving the ONIX parsing pipeline, reworking the Terraform and introducing reusable modules, and architecting both the data pipeline and the overall project structure. I also helped introduce semantic search and enabled a lot of features by building the Amazon integration.

I concieved and led the implementation of the search-to-product matching pipeline behind the discovery experience. It takes roughly 4 million search terms drawn from three marketplaces and matches them against half a million products, combining semantic search with an LLM-as-a-judge refinement stage to rank candidates. The result was a substantial lift in product visibility.

Separately, I built a scraping service that runs across a fleet of machines. It sends around 2.5 million requests to gather roughly 250k datapoints a day, landing in BigQuery and Elasticsearch for downstream analysis and retrieval.

Amplify automates marketing content generation for product catalogs. It reads product metadata and produces both copy and visuals at scale, turning a catalog into ready-to-use campaign assets without a manual design pass per item.

I discovered the opportunity while pushing for the Amazon integration, presented the idea, and co-led the project from initial concept to production ready.

It generates marketing copy such as setting hooks, quotes and hero lines, and lays them out onto banners. It also produces comparison charts, using semantic search and LLM filtering to pick the products and attributes worth putting side by side. Generated images run through an evaluation stage that scores and filters them, so only on-brand, high-quality assets reach the output.

A supporting AI-generated stock image gallery was built as a shared, reusable pool of visuals, giving campaigns a consistent set of assets to draw from instead of regenerating from scratch each time.

The Amazon integration closes the loop. It fetches existing marketing content and metadata, and pushes the generated content back to listings.

Audify is audio book and audio dramatization production software, built as a desktop DAW with cloud sync. It relies heavily on ElevenLabs for narration and sound effect generation. It produces straight narration from a manuscript as well as full dramatizations.

It runs a full analysis of the manuscript, identifying the characters and their traits and, from those, proposing a voice for each one along with narration tags, the emotion cues ElevenLabs supports, to fine-tune how each individual line is delivered.

It also analyzes scenes and their surroundings and generates audio effect settings that can be applied per scene, things like delay, reverb, EQ and distortion, and the same applies down to individual character lines. From the manuscript descriptions it proposes matching sound effects, and it proposes and generates the background music for each scene.

I helped ideate it and, as an amateur musician, took ownership of the music, sound effect, and ambience generation and syncing, along with the post-processing of every generated sound so the mix sits together cleanly. Shifting the soundscape to the story's surroundings this way makes the experience immersive and realistic.

The desktop DAW lets users manipulate each audio sample on a timeline and regenerate any piece of audio on the fly, with cloud sync keeping projects available across machines.

Working on the manuscript analysis pipeline, the DAW, and the audio generation provider integrations end to end lifted my understanding of how to build complex AI-powered text analysis tools, and took my audio production knowledge to a new level.

The need for a unified cloud platform arose organically across five business units. I drove the architecture end to end, drawing the schemas, leading the design meetings, and getting the solution reviewed by Microsoft's official Azure architects, first their general architects for the overall shape, then domain specialists for networking, database and compute.

The estate was large and fragmented. 30 microservices and applications spread across five teams, with multiple databases and their own deployments, ranging from on-premise to different cloud providers. Most handled complex construction element calculations, some served catalogue data, and others were auxiliary services such as PDF and XLSX generation. Bringing them onto one platform meant standardizing how every team built, shipped and ran their services.

I implemented the CI/CD and the Terraform that scaffolded the cloud services, introducing reusable modules so each team could deploy against a consistent, reviewed baseline. I dockerized and helped dockerize applications across the five teams and reviewed the deployment process throughout the migration. We used a micro-frontend setup where a parent shell pulls in smaller frontends that deploy alongside their respective apps, with Firebase handling authentication.

Partway through, a corporation-wide decision meant rebuilding the Azure solution on AWS. I ran the same review process with AWS's architects, and the groundwork from the Azure design made the second pass faster and sharper.

The result was a shared cloud solution deployed across two availability zones: Firebase for auth, CloudFront and WAF at the edge, API Gateway fronting Lambda and Fargate for compute, and Aurora and MongoDB for data, all behind a Palo Alto firewall.

The payoff was stability and uniformity across the units on a single baseline, lower cost from serverless compute that scaled down to working hours, and reliability from database replication on top of the multi-AZ layout. Security ran deep, adding GuardDuty, KMS, Inspector and CloudTrail on top of the edge and firewall protections already in place.

Leafworks is a Zendesk Premier Partner whose platform, Knots, extends Zendesk with automation and integrations, connecting it to outside systems like Shopify, ERPs, CRMs and Microsoft Teams for the cases Zendesk itself doesn't support. It started from a need for a no-code way to handle exactly those cases.

I drew the architecture, made the case for how it could work, and implemented the first version. It is a node-based visual programming model where support teams drag and connect blocks to chain logic together without writing code.

For the cases that weren't straightforward to automate, I built Studio, which goes further and lets clients write and run their own JavaScript. The in-browser editor is built on Monaco, VS Code's editor, with autocomplete, types, and a runtime API that exposes what the platform can do.

To run that untrusted code safely, each call executes in an isolated Lambda and a separate process, with the Node standard library stripped of potentially dangerous calls.

I learned a lot building the execution model, the editor integration, templates, custom variables, secrets management, the Zendesk integration, and the CI/CD behind it, with the codebase organized as a Lerna monorepo.

Two years after launch it crossed the 1M ARR threshold, which I'm proud of.

Affix Space is a software development studio that I founded in 2024 after retiring The Codefather.

A lot of things have changed since then, and I have learned a lot. But the core values and principles remain the same.

Development is changing too, AI is shaping our day to day work and I am excited to be able to kickstart this journey helping an AI first startup scale up.

The motto, credo, and mantra of my professional work has always been "deliver". If we can have some fun along the way, even better.

Chaptr is an AI startup in the publishing space.

I help them architect and build innovative AI solutions while scaling their development team and infrastructure.

Beyond engineering, I contribute to shaping their culture and values as they grow. Working with a team that builds new POCs and prototypes in parallel to their core products keeps the work dynamic and rewarding.

The role involves hands-on work with cutting-edge AI platforms, OpenAI, Anthropic, Gemini, Veo, Grok, ElevenLabs and more. I'm bringing a lot of software and cloud architecture knowledge, and am privileged to gain a lot of AI knowledge and experience, as well as some friends.

I broke my record in requests sent building a scraping service, roughly 2.5 million a day to gather around 250k datapoints. I helped automate audio book production. I proposed and co-developed automated marketing content generation software, partnering with a multidisciplinary team to design and deliver the solution.

A between jobs gig turned out to be a compelling project at the intersection of VR and construction. I served as the primary technical lead on Leviat's side, overseeing the development and delivery of an innovative VR proof-of-concept built by Unity, which I tinkered with a lot in the past, right as the company was being acquired by Capgemini.

The engagement provided a valuable opportunity to collaborate with exceptional engineers and gain deep insight into Unity internals, their VR tooling, and the processes behind delivering immersive experiences for the civil engineering space.

I have been tinkering with Unity for a long time, building small projects and games just for fun, so having the opportunity to work with a professional team and deliver a real product using Unity was a great experience.

As an owner of a one man software development studio established as a trade, I take on a wide range of projects, from small to large, from simple to complex.

Sometimes leading, sometimes following, sometimes doing both, sometimes just doing the coding.

It was a great experience and after 9 years it was time to retire The Codefather, and evolve into Affix Space.

Design and development of cloud solutions (AWS, Azure) using Terraform and CI/CD.

Development of TypeScript (Node, React, React Native, Angular, Vue), Python (FastAPI, Flask, Django) and PHP (Laravel, Slim, WordPress) applications, including the design and development of REST and GraphQL APIs and database schemas (PostgreSQL, MySQL, MongoDB).

Managed and led the design and migration to a unified cloud solution on AWS and Azure, replatforming over 30 applications alongside official AWS and Azure architects (covered under "Leviat Cloud Architecture" in Recent projects).

Consulted and led the development of multiple advanced expert systems within the construction industry, using Python (Flask, FastAPI) and Vue.js, supported by GitHub CI/CD and hosted on AWS and Azure.

Developed a fullstack TypeScript (Strapi, React, PostgreSQL) application with Stripe and DocuSign integration for an apartment leasing company.

Helped design and develop a serverless platform (AWS Lambda, DynamoDB, Aurora, Fargate, SAM, TypeScript, React, PostgreSQL) used by Zendesk applications with features such as advanced text parsing and OCR.

Designed and built Knots Studio, an in-browser IDE for writing code, along with the backend for safely executing code on AWS Lambda (covered under "Knots Studio" in Recent projects).

Architecture, development and deployment of a travel agency platform for a booking.com startup contest winner.

Designed and led the implementation of the architecture for a scalable cloud microservices application (Golang, Redis, Cassandra, k8s) for streaming video/audio from IP cameras (RTSP) to Ionic App (WebRTC).

Developed an asset management SaaS (MEAN stack, React Native) used by large construction works companies.

Designed and developed a specialized distributed text analysis software (Flask, PostgreSQL, Elasticsearch, React).

Developed webshops and web applications (Django, jQuery, Angular, React).

Worked on a variety of customer-specific projects ranging from Laravel projects, WordPress websites, Angular.js SPAs as well as the design and implementation of an in-house CSS/JS framework used on multiple Java projects and platforms.

Studio 42 was a digital agency that I founded in 2010. I was the owner, designer and fullstack developer, while still being a student.

I was working on a variety of small web projects, designing, developing and implementing them. It was a great experience and a lot of fun, but still I had the priority to focus on my studies, not growing the business.

As a solo dev in a Business guy & Tech guy team, I designed and developed a SaaS startup built with Angular and Slim PHP. The project was sold to an investment partnership.

Designed and coded, https://kacaveeda.hr/, https://mbtranslating.com, https://www.skroz.org/, https://www.metapiga.org/, https://www.referada.hr/, https://kulturistra.hr/, https://uniline.hr, https://zuc-vz.hr/, and many more websites that are unfortunately no longer live.

Delivered a dozen fully responsive React and PHP applications for CMG, some with PWA requirements.

Created an image cropping tool that is used by thousands of people around the world.

"Gave up" the maintenance of the project to another dude that at the time had more time on his hands.

Read more here

Created a Python testing utility that helped solve some testing problems my team was facing.

Created a philosophy encyclopedia that is a collection of philosophy books and articles.

It is a work in progress and I am adding more books and articles as I find them.

A passion project with no real monetization strategy, it's been taking forever because it is easy to get distracted and start reading the books and articles more than working on the project :)