Many of the possible project below involve SigmaOS, because that is what I am thinking about, but it is not a requirement to use SigmaOS (or use any suggestion from the list). Some of the suggestions can probably be morphed into a SigmaOS-independent direction. Most of the suggestions below are not well-defined projects; they are a starting point for brainstorming.

• Paper focused:
  • Take any from the papers and implement it yourself, or measure/extend the authors' implementation in interesting ways.
• Application focused:
  • build/port Ray-like system to SigmaOS
  • build a gg-like system for SigmaOS
  • implement ExoFlow in SigmaOS's MapReduce library
• Infrastructure focused:
  • implement WASM-based procs
  • support Python in SigmaOS
  • support many named's backed by one (or more) etcd's
  • use image-less containers to speed up Kubernetes
  • make SigmaOS run across clouds
  • make a vector-db cache
  • migrate procs
• Fault-tolerance focused:
  • extend kvd's replicated state-machine library with blocking ops
  • make kvd store its state persistently
  • chain replicate kvd shards instead of using etcd's RAFT
  • make state of microservices fault tolerant
• Security focused:
  • fuzz the proc-container interface
  • make a proc escape its realm in an interesting way
  • avoid running docker service as root
• Utilization/performance focused:
  • Measure/study cgroup-based performance isolation
  • Measure/study Kubernetes controler and make it faster
  • Explore using swapping for BE and LC procs to share memory
  • Make kvd elastic with load changes
  • Use bidding/spot market for realms
• Hardware focused:
  • Use RDMA to speed up, for example, non-linearizable reads in etcd
  • Allow procs to use GPUs