Data Science Consultant
Trekkaro
مجموع سنوات الخبرة :7 years, 1 أشهر
Implemented Data warehouse to make use of data for getting business insight and using the prepared data warehouse for email marketing and customer engagement and building visualizations for business insights.
Customer Segmentation and understanding of Customer Behaviour, Segmenting Customers using Clustering Algorithms (K-means, Hierarchical). Then reporting similar features for each group or segment such as frequently bought items, location, seasonal preference of treks etc.
Providing insights to advertising and marketing team to make effective utilization of marketing and promotional campaigns.
Developing automated Analytics pipeline to combine different sources of data such as trading/dealing department, operations, and marketing for MT4/MT5 reports.
Setup PowerBI dashboards for management reports.
I used Python to develop the data collection and analysis ETL pipelines for near real-time reports generation.
Worked with TradeButler a start-up based in Denmark, Project involves Pattern recognition of interesting regions of price movement, regions are breakouts of price, take-off of price etc. building classification models to predict a region (breakout/take-off) and trigger trading signals as per the strategy.
With Ster Innovation, I worked on developing a High speed trend predictive strategy trading algorithm for Bitvavo Exchange first to screen crypto currencies and then place orders for level 2 to level 4 depth order book data.
For Hazelpond Capital, I Develop Custom Indicators and strategies for QuantConnect platform. Developed 15 Pine Script (Trading View) Indicators and Strategies.
Position: Data Scientist
Project/Product: At Professional Traders we are into Forex Market Trading. I am working in the Research and Development team, responsible for creating Machine Learning and Deep Learning Models and deploying them on Flask RESTful API server.
Tasks: I was responsible for extracting historical data from MySQL database, creating features on pips based and time-based bars basically OHLC data. Analysis of data for feature extraction and data preparation, for predicting patterns that come up from financial indicators as well as trading indicators which was developed in house. Then training Neural Network models like CNN, LSTM, MLP and boosting algorithms.
- Developing python flask-restful web server API so that these models can be deployed and used in real time trading.
- Training Models using Scikit-Learn, Keras and Tensorflow. Experiments used to be done on Keras and Production Model trained in TensorFlow.
- Consuming Kafka server real time streams to build a pipeline that takes data from kafka-server and calculates triggers using model and then puts a trade on FXCM, through REST API interface.
Technologies: Python, Flask-Restful, Keras, TensorFlow, kafka-python, MySQL Database, Scikit-learn, numpy, pandas, Forexconnect, fxcmpy
Image Classification Model Building using CNN-Tensor-Flow, Keras for Classifying images of different objects available on the ecommerce website-the use case is to build a model for an ecommerce mobile app which captures image of an object and tries to give recommendation to most similar products.
My major project is Bosch Connected Industries, where Data is collected from sensors and analyzed for trend detection, change point detection, frequency detection for Predictive Maintenance. Also Includes Wavelet Analysis, Signal Filtering, Fourier Transform, Time Series Forecasting - Tensor Flow (LSTM), Anomaly Detection in Time Series IoT data.
Image Classification for detecting faulty welding objects images on welding production line in manufacturing units.
Masters in Structural Engineering.