Been mentoring a few engineers more formally over the last six months. Some things that have actually worked:
Positive reinforcement is underused. When someone does something well - catches an edge case in review, asks a clarifying question that changes the design, writes a runbook that future-them will thank - say so specifically. Not “great job”, but “the way you traced that back to the SLA requirement before writing a line of code is exactly the right instinct”. Specific praise is information. It tells people what to keep doing.
The other thing that works is teaching through questions rather than answers. When someone brings a problem, the instinct is to solve it for them. It’s faster, it feels helpful. But “what have you already ruled out?” and “what does the data say?” build something that handing over a solution doesn’t. People remember the reasoning they worked through themselves far better than the solution someone handed them.
Knowledge sharing works best when it’s attached to real work. A brown bag nobody asked for competes with every other thing in someone’s week. The same content, five minutes during a PR review or a postmortem, lands differently because context is already loaded.
None of this is complicated. It’s mostly just paying attention and being deliberate about where you spend your words.