What are the times for each one of the steps
1 - teradata to staging
2 - staging to final db
3 - final db to tabular (load data)
4 - tabular recalc (measures and calculated columns)

3 and 4 assume you did split the steps - if doing a processfull the only way to get the different values is through a trace.

- staging to prod db
Are the loads from staging to final db full loads or incremental? 
And are you dropping/recreating the indexes before/after load?
Is load using fast load

- prod db to tabular
is it on the same server or different servers - if different have you changed connection string packet size to the max (32768/7)?

What are the hardware configuration of both SQL and SSAS servers - how much memory allocated to each, number of cores, max DOP on the sql instance and so on.

If bottleneck is SSAS do you know if it is paging while doing the processing?
If bottleneck is retrieving the data from the view (you can test this with a straight ssis package to issue the same sql statements your tabular model is issuing and output to a file) - this would be standard SQL performance validation.

you may need to get the scripts from http://byobi.com/2015/10/powershell-scripts-for-collecting-ssas-related-perfmon-and-xevent-trace-data/
to get the perfmon and ssas counters and see where the issues are if on the tabular side.
My version attached including 2 scripts to load the output of perfmon/xtrace onto sql server tables. - and 1 to kick off both monitors in parallel 🙂