# Work (syntensor)

### Maksym
* Deliverables
* Serving 1M Graph ML model example
* Mixed instance types server with training script log wandb
* Front end set up server/back server + auth tokens
* Todos
* Start shared github for AWS building
* Figure out why boto3 is not booking AWS instances of some types
* figure out how to build mixed instance types cluster
* Implement bakend with mixed instance types + how to scale with mixed instance types
* Figure out data sync across instances
* set up FastAPI; and user-facing server
* Write script to book multiple instance types at the same time + run the python script in each to determine cost/result
* Training ML regression large graph example with wandb

### Clayton
* Deliverables
* Small/Fast version of model
* Full Model
* Instance optimized sub-component list
* Contact Ben Laurie
* ToDo
* Data Refactor -- Incl. Recon3D, UniProt, Reactome, Rhea, and the LINCS1000 dataset
* Breakout Structural Encoder Networks -> Length 64
* Provide quick/small version of the GDE model for setup testing (liver only)
* Provide alpha version of the model for setup testing (all tissues)
* Work with Michael & Stefano to determine the optimal instance for each sub-model
* WandB Summary statistics CI checking

### Zdenek

* Deliverables
* API Spec
* Auth using API Key
* ToDo
* Work with Maksym to set up AWS-based user authentication interface with both back-end and FastAPI
* Front-End User auth

# Notes
Security Consultant: Ben Laurie (formerly head of security at deepmind health): [e-mail]