NorthWest Biotherapeutics Inc (NWBO)

[brazen22]

can't believe you're doubling down with your completely inaccurate description of the Flaskworks device. I'll repost again for your convivence..
-------------------------
The published info on Flaskworks/EDEN shows a slow, deterministic perfusion system (µL/min flows) that drives a peristaltic pump, a few valves, and basic sensors in a closed cartridge. That’s classic microcontroller/PLC territory—no exotic compute needed.

Why an MCU/PLC is sufficient

Process dynamics are slow: set/hold a low pump rate, open/close valves, read temp/pressure/level at Hz-level sampling.

I/O is simple: GPIO, a few ADC channels, and I²C/SPI/serial cover it.

Data volume is tiny: timestamps + setpoints + sensor logs; standard Ethernet/Modbus/OPC-UA is plenty.

What the primary sources actually show

Patent US 10,647,954 B1 (Flaskworks/NWBO): octagonal chamber with 8 perimeter inlets + 1 central outlet, uniform perfusion, pumps/valves/sensors described at a high level—no need case for FPGA/PLD or on-instrument AI.

2018 Cell & Gene Therapy Insights article on EDEN: functional prototype uses a commercial peristaltic pump with ~8 µL/min per inlet and standard tubing/stopcocks—again, straightforward controls.

Why “AI optimizing each batch” doesn’t fit GMP reality

In GMP manufacturing, you run validated, locked recipes. Any change to control logic/setpoints requires change control and re-validation—you don’t let a learning algorithm “adapt” live on a patient batch.

If there are models, they’re typically pre-validated (PAT/QbD) with tight limits—not open-ended self-tuning.

Reconfigurable PLDs/remote logic updates = validation burden

Remotely reconfiguring hardware logic would trigger design controls, software validation, and re-qualification. For a device whose job is essentially “slow, steady perfusion,” that’s needless complexity and regulatory friction.

On “high-speed communications” and CMMI

The workload is logging and recipe download—not high bandwidth. “High-speed comms” adds no practical value here.

CMMI level is a business maturity model, not a regulatory requirement. Regulators care about cGMP/21 CFR Part 11/Annex 1/data integrity (ALCOA+), validation, and change control.

Bottom line
Everything public points to simple, robust control (MCU/PLC), not FPGA/AI. If there’s a primary source (device manual, patent claims/figures, regulatory filing) that explicitly specifies FPGA/PLD or runtime AI optimization, please share it—happy to revise if new evidence exists.