I got 4 years ML working experiences in Internet & Financial related fields.
Personally, I think the general working flow would be like this:
1. know what your custom want and know the data they got (maybe need to combine external data resources)
2. Store the data(sql,nosql,hdfs?)
3. doing ETL
4. feature engineering & featuring selection
5. model training
6. model evaluation
7. model deployment
8. model update(depends)
skills in pandas,numpy,scikit-learn, spark,hadoop, etc