Curriculum Vitae

Maintenance and grouping of in-house multiprecision libraries Rational, MpIeee and an exact plotting library of Jeff Tupper into our own GUI and programming IDE dubbed MPL (former Arithmos).

My own implementation of matlab using n-tree traversals and simple garbage collection using smart-pointers (nowadays a lot of that work would be easier using LLVM) was at that time ahead of the competition in matlab compatibility. But because it had to remain closed source due to the nature of my PhD grant or rather the decisions made by my research group, I was pretty much on my own to to remain ahead.

This did improve my knowledge of how to write cross platform interpreters and compilers and c++ a lot, but it became apparent we needed to go open source or the project would perish. Seventy hour coding weeks got me far but in the end could not match the workload that an army of volunteers brings to an open source project like GNU Octave brought to the table. Inevitably a few years later this resulted in MPL getting behind in completed features.

Also some teaching was done by myself: practica for C.S. Artificial Intellegence and Software Engineering (testing) which were mostly given using java and c++.

Build various in-house applications and components for webshops, dynamic dns for CCTV. Rails embedded/arm projects like web-based thermostat for iphone (way back when there were no commercial ones to buy). Also prototyped, patented and now produced + selling usb hardware key called AnyKey. Firmware was re-written various times in c++ and webshop was done inhouse (https://www.anykey.shop).

Also do smaller pet projects like 3D printer control using nodejs before octoprint existed. Play with various new languages and frameworks for self educational purposes. For instance now doing some crystal-lang work to see if it has future. In the past experimented with other javascript frameworks like meteor before settling on Vue.js and AngularJS as my 2 favourites. Ofcourse plain vanilla javascript and jquery still have their places too (it all depends on what is required to be done and how the other developers choices and progress has been in the past within a project. Good example is webturtle done in 2020 during my holidays as pet hobby project: it's pretty much mostly vanilla js that calls some phaser.js api calls. It's done without jquery or angular or react. Except for the code editor used which has a dependency on jquery, it pretty much runs on all modern browsers with plain vanilla js. I also run my own dedicated linux servers on hetzner since the early 2000's.

VIU2: a new 3G security camera with linux based firmware. For this project I did some kernel driver c embedded work to implement image flip 180 degrees with 20fps h264 hardware. Further did a lot of server side work. They initially hired me to improve a buggy php project. But luckily I got the chance to convert it all to ruby on rails by promising a rewrite in the same time they estimated the bugfix period. I met my deadline and they were so impressed I was allowed to use ruby for various other projects at Option NV. We did a JSON api with ruby Sinatra for android and iPhone apps of Viu2. OTA firmware updates and also other ruby on rails projects for managing the VIU2 cameras in the wild.

Xyfi: Worlds smallest wifi router/hotspot in a usb stick. Here I was employed to get the web interface ready for release (jQuery GUI, LightHttpd and cgi app on OpenWRT firmware on the key).

APM: Ambitious project with AT&T to do traffic monitoring and traffic shaping by running software embedded on each android device. About 2 million devices in the field. Server side we used python twisted due to reactor pattern (basically what nodejs/express is all about these nowadays). We had 8 dedicated servers and a machine learning app on the device side to allow traffic shaping to be done on the AT&T network).

Also wrote a custom angularjs directive (widget/gui component) to do the calendar planning of badgecontrol.com and various other smaller projects whenever Rombit needed it. For instance we did the GAS fine app (gasboetes) in 2 weeks that connected with a ruby on rails json api (I wrote and was able to introduce rails for this pet project but due to other developers favoring php/symfony it was unfortunately not adopted in other backends). Even though it proved faster in both development time and in performance compared to their existing php symfony components. I heard somewhere they switched to java spring by now though ;).

Short assignment of 2-3 weeks with old colleague from University of Antwerp (Pieter Van Gorp). He's now professor at TUE/e (University of Eindhoven) and needed to get his backend for a project called GameBus developed. He wanted the implementation in java so I set up 20 resources using Java Spring and connected to postgres using Hibernate ORM. Using lombok for code generation this was pretty efficient. For authentication/jwt token generation I used a small rails app to provide a login screen during the last day (due to time constraint I reverted to my favourite framework ;) ).

Short assignment of few weeks to study/prototype a raspberry pi and see if it could replace the current video conferencing hardware used. Got to a working version with h264 codec but they had settled on the (inferiour) mjpeg coded for transferring video. My opinion was that they could and should switch the other components to h264 also. They had other congestion issues arising from sticking to the antiquated motion jpeg video format. Furthermore using mjpeg on raspberry pi was slow/choppy with video playback, even on an utp connection. Whereas the h264 with the dedicated decoder allowed butter smooth 25fps video over regular wifi. So basic conclusion mjpeg on the Pi2 was pretty much a no-go unless you'd spend (or rather waste) a lot of development time in optimising for an antiquated codec.

In a nutshell: this allows for large scale CFD computations to be done using Google Compute Engine virtual machines and the only limitation to processing power is the amount of funds available to pay GCE instances. The web app also had the usual features like keeping track of all users projects, users and user credits for this system and auto email PDF reports from aerodynamic simulation results. Users can upload 3D stl files, view them realtime in a viewer and then select options for the simulation type (set windspeed etc.) after the calculation on a seperately started GCE this GCE is auto shut down and the results are downloaded to the main server and also presented as a PDF summary report that is emailed to the users. (All work was done in 4 months and just 3 days per week for writing the whole system from scratch to a working prototype with all functionality and 4 project workflow redesigns along the way ;) ). It did need some further css styling and finetuning of the config files for higher loads, more automated tests and most likely moving some more stuff to background tasks. But the prototype was delivered and working completely: from uploading a basic stl to a finished simulation result PDF including all above described features and with working email notifications.

Worked as senior/lead developer in ETS group at Engie. Communicate with various api's for aggregating windmill and other renewables data from engie. Arkona, Metologica, Day ahead include some of the api's used. Various workers implemented in something called 'octopus project'. Also wrote some screen scraping for migrating data from pipedrive to a different system in python using a headless chrome extension to get into the pipedrive systems because the api did not expose all data (extracting emails, attachments and deals json data). Various improvements in large flask based application called b2e2 (back2earth) for managing energie contracts, sites and servicepoints.

Visual Graph generation of database structure using pygraphviz made to work inline in b2e2. Maintenance and curtailment management using vue.js frontend and flask rest-less api. Some standalone clients written like meteologica- client.

On frontend side did some Vue.js/nodejs frontend work in this context, optimising various pages and creating some overview graphs for maintenance events on etc.

Very strict development using TDD, code reviews and continuous integration.

Indexer: A ruby on rails application that can be configured to automatically sync certain tables to an Elastic Search index. This iterates and monitors all data from video and audio archives in a postgres database. Using some api calls a batch re-index of all data can be done. Also we have postgres triggers in place with skor daemon to allow changes to propagate to elastic search after eache update/delete/create done in database on various tables. Backend in ruby on rails and sidekiq workers for larger background jobs but also for the triggers for better performance (if a lot of triggers come in the webhook gets response instantly and redis queue + sidekiq handles the actual work). In the more recent version of hasura there are now fully supported webhooks that we will probably use to replace the current postgres triggers for better roboustness. Sidekiq workers make this nicely scaleable and efficient.

GraphQL jwt authorization app: With a login page and a minimal ruby sinatra application we implemented authorization and authentication for users on the GraphQL api.

Hasura GraphQL: complete and configurable graphql app with migrations for providing the api calls for frontend team developing the websites. Work was done to integrate with the custom jwt authorization app and made docker containers for all above to deploy with OpenShift and/or any desired webserver. Also CI/CD and automated tests written and gitlab builds docker images of the 3 components.

S3 proxy: Ruby sinatra app that proxies Caringo S3 server requests to add additional functionality to the S3 protocol and allow archiving to be done and connect to in-house api's to keep track of archiving status and caching of large files exchanged over S3 for various content partners like the public broadcaster VRT.

LDAP-SSUM: Rails app that uses devise for login and registration forms and CSRF tokens. For account storage it connects to an inhouse OS LDAP server. With reCaptcha integegration to keep out unwanted registrations. This registration system is being used for hetarchief.be, and soon the new avo site used in education/schools all over belgium. Most likely many more sites in the near future will use the same pages to provide an easy way to have accounts working across different platforms. In a nutshell it's a bit like accounts.google.be but then tailored to Meemoo's needs.

VRT metadata micro services: Micro services written in python with flask. One lightweight dispatcher service connects to rabbit mq which exchanges data to and from VRT for archiving media content and its metadata. This dispatcher communicates with various other micro flask apps that process the xml data and transform it into json to be stored at Meemoo's data centers.

More projects mentioned and added here Meemoo customer page