Find my blog on blog.londogard.com!
And make sure to not miss github.com/londogard where I do a lot of code today!
Recent entries:
- Timeseries (predicting stocks/crypto) 3-part series (#1, #2, #3)
- Deep Learning in the browser using ONNX with KotlinJS
- Knowledge Distillation
Almedalsveckan 2021 - Vikten av att verkligen förstå AI-modeller (The Importance of Really Understanding AI-models). I did a presentation on the importance of understanding models and their decisions. This talk touched upon an array of matters such as Ethical AI, Explainable AI (XAI) and Data Bias. The focus of the presentation was to be of societal contribution.
There was multiple people showing up afterwards thanking me for a great "easily digestible" presentation, one being Hallands Län.
youtube.com
Industry Days 2021
Based on my expertise I was invited as an AI Expert on a panel to discuss the industry of tomorrow and how to speed up the digitalization during Industry Days 2021 (by IEEE) in Västerås.
vimeo/ldlmedia (starts at ~2h13m)
Google DevFest by GDG - Managing the ML lifecycle without a headache (MLOps)
I did a presentation in person that unfortunately didn’t get recorded as the camera battery died. I did a reshoot but it didn’t become as natural as there was no audience. Either way I was speaking about the importance and how to manage the ML lifecycle without a headache. I presented some great tech agnostic MLOps tool (DVC/CML) and how to work with these.
youtube.com
| What | Where | Summary |
|---|---|---|
| londogard-nlp-toolkit | Kotlin, JVM | One of the, if not the, best Natural Language Processing Toolkit on the JVM written in Kotlin. Includes simple usage of Tokenizers (including HuggingFace), BPEmb, classifiers, stopwords, and much more! |
| Summarize | Kotlin, JVM | A summarize with two different variants which were state-of-the-art for extractive summarisation a few years ago when I implemented them. One is built on top of TFIDF, the other combines TFIDF, embeddings and more! Runnable on the Raspberry Pi by using some clever optimisations for Word Embeddings (try it here) |
| Text Gen | Kotlin, JVM | Text Generation without Machine Learning. Built purely on statistics, clever optimisations and smart lookups allows this too run on a Raspberry Pi |
| Java SymSpell | Java | A port of SymSpell to Java & the JVM. SymSpell is a incredibly efficient spell-correction tool |
| Fuzzy Match | Kotlin, JVM | A copy-and-insert ready to use library for fuzzy searching, works incredibly well (try it here) |
Make sure to take a look at github.com/londogard, londogard.com & blog.londogard.com.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent