This is an interesting question. You could use differents strategies. First you could decorate or functionalize nano particles with flame retardant molecules and then mixing by extrusion. Also you could try to add a coupling agent to "bond" these functionalized nano particles wint the polymer matrix. Reactive extrusion is also used to help this reaction.
Without specific knowledge on the topic of flame retardants, I would say that the common practices to incorporate nanoparticles into polymeric structures involve mixing (i,e,n in compounders or extruders) and surface grafting strategies (usually in situ polymerization). The first approach usually entails problems associated with low miscibility, poor dispersion and possible degradation phenomena and may be accompanied by the use of specific compatibilizers (as mentioned in the previous answer).
Maybe, but then probably other chemical or structural properties of the polymer would change, and probably you don't want that. You would have to check also the chemical feasibility of doing the synthesis of a copolymer (either random or block for example) composed of "retardant" monomers and "normal" monomers, and I say monomers, not atoms to be more precise. If the "retardant" monomers are not able chemically (in terms of reactivity for example) to compose the chain, then you just can add the retardant later as in reactive extrusion or a post-grafting process as suggested in previous answers.