In his quest to build tumor killing bacteria, he articulated some steps useful for building synthetic biological systems.
1) Define spatial and temporal environments that the bacteria encounter.
2) Identify processes that satisfy the environmental constraints and promote environmental progression.
3) Identify modules that satisfy these requirements.
4) Build a minimal system from parts.
5) If required, go back to 2), re-build and add complexity.
Of course, unless the chassis has no system that is already running, the systems that are being built here are done in the existence of a host system, which perform
1) DNA replication
4) DNA repair
5) Protein enzymatic functions
6) Protein non-enzymatic functions
7) Protein degradation
8) Cell division
9) Metabolism for energy
The above applies to the construction of synthetic system using plasmids, which is what we will likely use. Some potential problems with this approach is cross-talk of newly built system with the host system. The concept of orthogonality  should be understood here because it can be useful to avoid cross-talk. In terms of optimizing some parts for specific functions, the concept of directed evolution is powerful and was applied in the tumor killing bacteria example above. Of course, directed evolution is only possible if positive selection for the part functions is possible.
The generic approach presented for sensing, processing and actuating responses with bacteria can be tailored for some applications on top of my mind (see below). The study of these applications require our engineer friends to tell us how to study the transfer functions from sensing to processing, what logic devices (eg. Inverters, AND gate, switches, etc) to use for processing, and how to study the transfer function from processing to actuation. Of course, modeling at the level of device and system function will be appreciated too.
Can we built bacteria to interface with cells other than mammalian cells?
1) Perhaps insect cells
GFP as reporter on our bacteria invading different cell types. If germ cells can be invaded and transgene can be incorporated to the insects, this is a way of transforming them. A suicide gene may be installed, etc. Can a light sensor be coupled into the system to direct bacteria movement in vivo?
2) Can we built bacteria that interface with mammalian cells but do different things than killing tumor?
3) Can you use your imagination to think of others?
References: Bacteria that sense and move toward a chemical. riboswitch as sensor and processor, altered cellular behaviour as actuator