I'm pretty new to QuantConnect, but have been taking a deep dive for the last couple of weeks and prototyping a trading system.  I have a deep background of application development/BI with the Microsoft stack, with a relatively more recent focus on data science and machine learning, so the power and potential of QC are quite exciting to me.

However, the restrictions on whitelisting functionality of the .Net framework is of concern, and I'm trying to understand what the potential workarounds might be.  There are two main use cases I'm thinking about that I believe I'm currently restricted from implementing:

1.) Data persistance for the purpose of algorithm configuration optimization.  I want to be able to store a record for each backtest with its outcome, runtime, and user-defined configuration parameters.  I want to be able to easily query and report on these records.

2.) I want to be able to train predictive models in other tools (most likely R) and operationalize them so that my Lean algorithm can use them to make scoring decisions.  I believe the two most likely approaches for operationalizing an R model would be either use of R.Net or integration with SQL Server 2016.  If anyone has integrated a model with Lean another way, would love to hear your approach.

I can implement these use cases myself, but it will require utilization of non-whitelisted components of the .Net framework, which I understand means no use of QuantConnect for backtesting?  

If I do choose to go this route of running Lean myself for live trading, without use of the QuantConnect platform, is there anything else I should know?  Aside from QuantConnect's backtesting data and reporting UI, will I be missing any other core functionality?  Are there others that have gone this route with Interative Brokers that might want to compare notes as far issues and configurations?

Thanks, and looking forward to contributing to this community.