- SEEKING WORK | Frankfurt, Germany | REMOTE-only
I have some time on hands I can spend on Machine/Deep Learning in computer vision, NLP and fraud detection (spot instances and series). I am both in bleeding-edge academia as well as in industry productionalizing state-of-art models (i.e. replicating proven papers that are 3-6 months old and putting them into production on problems that largely reduce to their assumptions). Recently turned a DenseNet and Wide ResNet multicategorical image detection into an SaaS serverless product, prototyped a Spark-based "classical" ML fraud detector ensemble matching performance of a hand-crafted one (Gradient-boosted Decision Trees, Random Forests, SVM etc.), identifying suspicious features using PCA/t-SNE/clustering and preparing another detector based on spot fraud detector (VAE) and time series fraud detector (VRNN/WBW), e.g. bot traffic. Automating customer service using sentiment analysis and mail classification/response generation. Self-driving cars, Deep Reinforcement Learning for robotics and program synthesis for research fun too.
Purely project based, i.e. you specify what you need and I work on a solution with agreed checkpoints (renumeration possible based on days or progress level).
While based in EU, there is no problem to spawn a company in Delaware to service US-based clients if needed.
If you are interested, shoot me a mail to peter.skvarenina at protonmail.com
- Location: Frankfurt, Germany
Remote: Remote-only
Willing to relocate: NO
Technologies: Distributed Deep Learning, DApps
Résumé/CV: upon request
Email: peter.skvarenina@protonmail.com
I am offering remote freelancing in the area of the very latest distributed Deep Learning and/or DApps. I have experience programming state-of-art 2018 Deep Learning computer vision classification models (DenseNet(-BC), Wide ResNet etc.), reproducing world-class results, running distributed training and hyperparameter optimization (AWS Sage Maker, tf.Estimator), productionalizing Deep Learning microservices (AWS Lambda - Chalice/Serverless, GCF, minikube/kubeless/K8s etc.). Worked on e-commerce DApp prototype in Solidity (no ICO offers please).
World's top-10 US-based college education in Machine Learning/Robotics; nanodegrees from Udacity in Self-driving car, robotics, AI and Deep Learning. Worked for some of the world's best engineering companies already.
Working either on an EU-based contract or spawning a Delaware-based corp for US clients.
I am not cheap and I am good; no-nonsense straight-to-the-point approach, hands-on with low-level details, self-driven, no tech guidance needed, fully accommodating business needs; expecting reliability back.
- Location: Frankfurt/Germany
Remote: Remote-only
Willing to relocate: No (for the next 6-12 months)
I am a developer working in 60+ languages (Java, C++, Python, Scala, Go etc.) with a huge range of skills such as Deep/Machine Learning, Cryptology, Computer Graphics (3D/2D/geometry), Robotics, Computer Vision, NLP, Speech Recognition, Compilers, Big Data, e-commerce, Android, currently learning AR/VR/iOS and e-commerce DApps.
World-class education (top 10 engineering school) as well as working experience.
I've programmed an own SQL-interpreter, vector animation software (www.animatron.com), Google Earth-alike viewer for 3D cities, 3D engine for railroad simulation, fully automated e-commerce SaaS, online multi-language dictionary, car racing timing software; worked on distributed enterprise messaging with high availability and transactions, machine learning system for health informatics and many more.
One of the first graduates of Udacity's Self-driving car, Artificial Intelligence, Deep Learning Foundations and Robotics nanodegrees.
I am looking for smaller high-energy companies for permanent or contractor hands-on positions (love to create new things together with motivated and non-complicated people).
See my CV with contact details at http://bit.ly/2nA1XpD and get in touch if you are interested!
For US-based positions/contracts I can make a personal US-based LLC used to handle all administrative tasks.
- Shameless plugin: Here is a recent presentation I did while finishing Udacity's Self-driving Car Nanodegree. It might help with some of the concepts, show you what was Udacity teaching and complement the material from MIT:
Have fun!
Location: Germany (Frankfurt am Main) Remote: Yes Willing to relocate: Yes, in 2-6 months (German citizenship pending) Technologies: Machine Learning, AI, Computer Vision, Big Data, Robotics, 40+ programming languages (ex-SUN, ex-NOKIA, now JetBrains) Résumé/CV: http://www.squared9.com/documents/cv.pdf Email: peter.skvarenina/at/gmail.com GitHub: https://github.com/squared9- Actually SMIL is not that interesting per se, its importance stems from being the only open standard for professional grade declarative vector animations supported by most browsers right now, i.e. equivalent to Flash.
As it's going to be removed soon, we would have to wait until Web Animations are finalized and properly implemented everywhere. Frankly, figuring out all the details of both specs and implementation deficiencies in browsers for SMIL wasn't the nicest experience (WebKit/Gecko/Blink), however there was a way to get it working in all of them in a single manner, and get 1:1 quality comparing to Flash animation. How long would it take for Web Animations to get there? As someone who was a member of a committee for an unrelated automotive standard as well as working with committee members of an enterprise messaging standard, I know very well how easily given various political interests this could turn into a solution nobody really wanted.
SMIL and IE have a complicated history. On a vast majority of recent mobile devices SMIL is perfectly usable (until Google drops the axe) and there was a surge of interest wrt SMIL driven by the availability of retina/HiDPI (small-form) displays and the need to scale animations properly, and we have an indication that our tool supporting SMIL was causing some waves in the industry as well.
I am also not sure why you'd assume only hobbyists and enthusiasts care about SMIL. I personally was dragged into it by the fact nothing else was anywhere near it functionality wise and it had to be implemented, while I could imagine much better standards myself. It was a completely pragmatic, non-enthusiastic reasoning.
- Obviously Web Animations look very good given they seem to solve the (IMO) worst weakness of SMIL - time manipulation. In SMIL to get the same effect you'd have to approximate local time by complex easings and cut/reverse parts of animation depending on the direction of local time. Web Animations took a much more reasonable approach (I was actually puzzled SMIL had such a flaw in its design).
However, as a producer of a professional animation tool I can't wait until Web Animations show up finally someday, but have to address the need expressed by customers that call for open vector animations working on all devices right now - and SMIL was the only one that could have address that. Now, there's nothing until web animations show up, and we can't be guaranteed that the final implementations will be anywhere near promised functionality.
Even just retaining SMIL in its current state would have been much better than just outright dropping it for any kind of continuity. We will obviously provide Web Animations as well, but with the deprecation of SMIL we lost a stop-gap declarative solution for majority of mobile devices, and frankly I don't understand such a haste in dropping it (just sustaining it in Chrome at its current state can't be very difficult)
- I am of the opinion that SMIL has been dropped prematurely - there is no real alternative for the time being, web animations aren't finalized nor supported yet, and just keeping SMIL in the current state without developing it any further would be still better than the other alternatives (CSS/JS).
- As somebody who spent a lot of time and effort to figure out how to do Flash-style complex animations fully in SVG SMIL including reversed time loops, hierarchical time modelling via easings and to make authoring available for free in an online animation tool, I am very sad to see abrupt Google's decision to deprecate SMIL in Chrome without having a full replacement available. SVG + CSS can be used only for simple things, SVG + JavaScript competes with canvas + JavaScript and more efficient geometry formats, and in addition in both cases the declarative approach is lost. The goal of having Web Animations to take place of SMIL makes sense, especially with advanced time modelling, however there are still many cases where the current proposal can't do what SMIL does and it would take some time until Web Animations will become usable in all browsers at the expected level for professional animation. SMIL has its share of problems but it was more-less working on all relevant mobile browsers. Now we are basically left with nothing that can do the job.
- Just a question - why don't you use SVG SMIL animation? It's perfectly fluid, HW accelerated, has declarative animation transforms, allows easings, draws in its local coordinate space, you can add interactive JavaScript events etc.
- The original Java code for bucket fill is very fast, sub-second fills in most cases however after converting the code to JavaScript the performance gets worse in some specific cases (Bezier offsetting/flattening combined with Boolean operations for non-closed or transparent shapes). There are some advanced math and algorithmic tricks used in the Lasso tool to speed up exactly the same class of problematic cases in JavaScript. Bucket fill is however more complex which doesn't allow straightforward reuse of these algorithms; additional algorithms are currently being developed to deal with it.
Good luck with your nice paint app! ;-)
https://drive.google.com/drive/folders/14DKPte-1YvU5IV3Rb8zd...
The issue back then was that Google suddenly announced their intent to deprecate and remove SVG-SMIL from Chrome, instantly freezing all development there. However YouTube complained and made them change their minds; later they put it into a "frozen" mode but didn't remove it.
For most artists it would have been a risky bet to use it if Google didn't want to support it any further, despite the promise of having an official standard for animation in all browsers.