Most of what I'd look for has been covered here already. Key things I look out for when combing through portfolios:
Use of different game engines - More often then not you'll be joining a studio and using either their own engine or a modified one. If you can show you can transfer your skills from one engine to another (thus reducing the amount of training time you'll need when joining a new studio) it'll be a big bonus. Basically, don't just make CS:GO maps.
Ability to work on different game types - This kinda relates to the above but if all you're making is CS:GO maps (sorry to pick on CS:GO :P) then you're probably not going to be considered for a puzzle/racing/adventure game potentially. If you show you can create levels for different game genres it'll really help you out. For example make a race track in Trackmania, then go create a Portal 2 puzzle level, maybe create a Fallout 4 dungeon then go create a single player mission for your favourite shooter with their in-built mission editor.
Using level layout to create different gameplay - It's nice to see some variety in level layout. Try something new in a CS:GO map which breaks away from the typical 3 lane structure. Maybe a level which predominantly focuses on height variation. Perhaps the Portal 2 map you create can focus on precision shooting and the timing of portals being placed. The Trackmania track you might create could use split/multiple paths to give players more choice rather than one set path?
Tell me what didn't work - A lot of the time what you create might not be the best level ever. And that's totally fine. But don't hide it away at the bottom of your portfolio or remove it entirely. If you can critique your own work, say what went wrong and what you might have done to fix it it'll show understanding and show employers how you might tackle problems during development.
Keep portfolios short and sweet - Kinda contradicts the above point but don't make a portfolio page about a particular level really long. Game devs are often busy people at the quietest of times so we don't have time to read essays on every aspect of your level. Show us the key gameplay features, maybe a quick level playthrough video and keep any text short and to the point. This normally helps with interviews too as it leaves us something to ask you. Like: "You touched upon reasons why you didn't like this section, could you expand on what you might do given the chance?" "You mention bombsite B was always hard to retake in playtests, how did you overcome that?"
We don't expect you to be artists - Don't spend forever trying to make your level look super amazing art-wise. You're not expected to be an environment artist too. So long as you've shown considerations for how the level might be dressed or themed then that's often enough. Mood boards are good and some basic dressing to help sell that in the whitebox is cool too. Resuse assets from games as much as you can.
I agree with what others have posted. If you don't know any 3D packages right now then just focus on in-engine tools. They're more than good enough to whitebox a level. But if you do know a 3D package it'll be a nice bonus on your CV. At Rebellion we let people use our in-engine tools or a 3D package (Max, Maya or Blender) when building their levels.
Give your portolfio variety
Don't waffle on too much (Like my post :D)