This is solved in my paper [here][1] using a generalized version of the techniques of Sapir.  We also show many other properties of the embedding can be achieved while adding malnormality, e.g the preservation of the decidability of the Word problem.


  [1]: https://arxiv.org/abs/2404.00841