Killian  .
BerlinGermany

AI & Machine Learning | Apache Hadoop, Apache Lucene, Apache Spark, C

? Selected Projects Dialect Classification Classify English dialects from six regions of America. Trained a speech-to-phoneme model and fine tuned on dialect data. Neural Chatbot for Technical Support Trained an end-to-end neural chatbot model on Ubuntu forum corpus. The basic archetecture is HRED (Hierarchical Recurrent Encoder-Decoder). Used adversarial training and future-aware decoding to alleviate the 'dull response' problem. N-gram Language Model Implemented a count-based tri-gram language model with Kneser-Nay smoothing with Java. Applied several optimizations in order to fit the model into memory while keeping running time reasonable. Used openaddress hash tables to build the model, and then sort and compress the tables for less memory usage and faster query time. Small-C Compiler A compiler front-end that translates small-C (uses a subset of C grammar) code into LLVM intermediate representation.

Skills

  • Machine Learning
  • Machine Learning Model
  • Apache Hadoop
  • Apache Lucene
  • Python
  • JavaScript
  • Java
  • Django