I can't give a simple answer ie. X millions died, but I can probably tell you why you'll have trouble finding a simple answer, and perhaps why we'll never have a straight answer to this question.

The basic issue you'll encounter is that some sources in reporting on 'how many people Stalin killed' are not all talking about the same thing. Some of those sources are looking at different time periods, be they narrowed down to the purge and famine years, or from a start date (exactly when you want to say Stalin was 100% in charge will depend a bit) through to 1953 when he dropped dead.

Then of course you have to define what counts as Stalin killing someone. Some sources only count executions, which are pretty clear cut if you want to explicitly point to people killed by the state. But others will include the Great Famine/Holodomor (culpability, how, why etc. for which is also a large area for debate), deaths in the Gulag system and so on. These end up in a grey area because there doesn't seem to be an indication that Stalin decreed that more people need to be worked to death at Kolyma for example... but decrees he did give certainly resulted in people dying. Also clouding this is that the 'numbers game' is often used for comparison purposes ie. 'did Stalin kill more people than Hitler', but should you compare the Gulags with Aktion Reinhard?.

The next problem is the lack of information. Estimates made during the Cold War were typically extrapolations based on population date, what official reports saw release and similar. Post Cold War there hasn't been quite the opening of the archives that were expected, a lot of material that would give you internal reports on mortality in the Gulags etc. are still not completely open (which again leads to extrapolation and guess work). Not all the information is straight forward and easy to find once you consider all the potential areas you have to cover (famines for example aren't really going to be detailed in NKVD files, executions are going to be filed separately to the national operations etc.).

Then you have classification problems - for example official deaths in the Gulags would have been deflated somewhat by a reasonably common practice of releasing inmates who were basically at deaths door early - so they die outside of the camp system and hence don't factor into reports. The notionally straight forward executions can be a problem because criminal offences like theft can by effectively political crimes. The Law of Spikelets for example - stealing collective farm property - could attact a death sentence... but is it just theft or do you class this as resistance, or a hungry people stealing food? On the topic of executions, do we perhaps really mean 'murder' when we say 'killed' in this context, or should every execution count? Trying to separate this all out would essentially require you to investigate every case individually, which isn't really feasible.

You've probably already seen this link around the place. It is a decent starting point to see the wide array of results different scholars have gotten in trying to estimate the death toll under Stalin. It also touches briefly on the big/small numbers 'debate'. In terms of going into more depth I could probably refer you to some texts (and academic journal articles if you're super keen) about specific areas like the purges and famines. I actually haven't got anything good about the 'national operations' which is an often neglected area in terms of Stalinist tyranny.