Tell us about the workload. We will come back with an architecture, a cost shape, and a straight answer on feasibility.
The more concrete the workload, the more useful the first reply will be.
Include model size, parallelism strategy, token budget, and target timeline — it lets us answer in one round instead of three.
Use the same address and mark the subject line accordingly.