Technology stack
Our data science team can perform wide range of tasks. They have strong knowledge at such solutions:
Technical qualifications
- Natural Language Processing (NLP): Data preparation, sentiment and similarity analysis, topic modeling, keywords extraction.
- Fine-tuning state of the art models: GPT-2, BART, BERT, ALBERT, DistilBERT, Grover
- Development of custom metrics for semantic analyses of generated text
- Deploying models as high load APIs on cloud infrastructure
- Computer vision (CV): Object detection, object classification, face recognition, emotion recognition, segmentation, pose recognition.
- Implementation of detection, classification, and recognition algorithms on static images and real-time video streaming data
Transfer learning
- State of the art models: MobileNetV3, VGG16/19, ResNet50, RefineNet
- Deep learning frameworks: Tensorflow, Keras
- Detection/recognition frameworks: SSD
- Machine Learning (ML): ensemble methods (XGBoost, Random Forest, LightGBM), model validation and training (boosting, cross-validation), hyperparameters fine-tuning, feature engineering
- Forecasting modeling: time series trend and seasonality modeling (Fbprophet, ARIMA, SARIMA)
- Front-end Data Science framework: Streamlit
- Programming languages and frameworks: Python, Django, Node.js
- Python data science libraries: NumPy, Pandas, SciPy, SKLearn, TensorFlow, spaCy
- Fast search ANN: Annoy.
