Leading all of the IT development at tech-enabled minibus transportation service in Kazakhstan with 1000+ vehicles under management and 3000+ daily trips

Developing Twitter's Graph Storage

Real-time data infrastructure. Highly scalable distributed streaming system.

Leading research and productionization of a group of backend software developers in the following areas:1. Acoustic modeling via Connectionist Temporal Classification technique.2. Language modeling via classical n-gram and RNN-based approaches.3. Attention-based g2p modeling.4. Embedded speech recognition.5. Phrase Spotting.

Improved previous results in phrase spotting by 10% of relative error. Deployed it in production.Started experiments in end-to-end speech recognition.

Got 20-30% of relative improvements in alignment text with audio compared to baseline statistical approach.Started experiments with phrase spotting.

Practical NLP lectures in Russian.Responsibilities: teaching more than 100 students, creating a syllabus, designing course goals and tasks, checking students assignments.

Improved confidence estimations in ASR systems by means of developing new techniques in this area. Got 25-50% of relative improvements.

Practical NLP lectures in Russian.Responsibilities: teaching more than 100 students, creating a syllabus, designing course goals and tasks, checking students assignments.

Made a course about Automatic Speech Recognition and Speech Synthesis for computer linguists.Topics:1. Basic Probability Theory2. Language Modeling3. Automatic Speech Recognition4. Speech Synthesis5. Phrase Spotting

Studied and applied Continuous Data Assimilation technique to the 2D Benard Convection Model.