The opportunities aren't hard to find — they're scattered and time-boxed

An information-structure problem

If you finish a master’s in experimental psychology in 2027 and want to move toward cognitive neuroscience, brain–computer interfaces or brain-inspired AI, here’s the map you face:

BAAI posts its joint-training PhD call on its own site. The Chinese Institute for Brain Research opens general admission around October. Tsinghua’s Department of Psychological and Cognitive Sciences runs two rounds, and the second only opens if the first doesn’t fill. Shanghai AI Lab co-trains with a dozen-plus universities, each with its own application system. And a whole tier of university research-assistant posts appears only on a lab’s own page or in an advisor’s social feed.

None of this is hard to find. Take any single item and ten minutes of searching turns it up.

The problem is that it’s scattered and time-boxed. Dozens of institutional sites, each on its own publishing rhythm, with windows that stay open a few weeks. You can’t patrol all of them daily, so most opportunities reach you after they’ve closed.

BRAIN / 27 Opportunity Radar (radar.openagent.hk) pulls that map onto one page.

But the list isn’t the scarce part

Having built the aggregation, I realised a complete list is less useful than it sounds.

Someone graduating in 2027 has all of these live at once: apply for a PhD, apply for joint training, find a research assistant post, join campus recruiting, take an internship. The windows overlap, the materials conflict, and most people’s real problem isn’t not knowing the options — it’s not knowing which one to do first.

So the radar doesn’t rank by institutional prestige. It ranks by whether something can produce a result in the next 30 days. The top of the page states a shortest path outright:

Apply first to university research-assistant posts open to master’s students, prioritising Beijing, Shanghai, Shenzhen and Hong Kong; accumulate data, papers and references while preparing the 2027 PhD applications in parallel.

The judgement underneath: PhD admission turns on research output, and research output takes time to accumulate. Carpet-bombing PhD programmes today means applying with nothing strong in hand; six to twelve months in a lab first means applying with a completely different set of cards.

What each entry carries

14 selected opportunities, each with:

• current state — open / expected 2026.10–11 / keep watching
• direction tags — brain-inspired intelligence, AI4Life, cognitive neuroscience, BCI, rotation-then-advisor…
• the next concrete action — down to the granularity of “write the 5,000-word research plan and 1,500-word statement of interest now, and line up two academic references”
• a link to the official source

That last one is the floor: the radar is navigation, and the binding information is always the official announcement.

You can filter by type: PhD / joint training / research assistant / campus recruiting / internship / research post. Beyond the list, the site carries an advisor radar, recent papers, a research-institution map, source list and collection log.

It’s a static snapshot

The collection date sits in the header. That’s deliberate.

A live radar scraping daily sounds better, but for this kind of information “looks live” is more dangerous than “clearly a snapshot”. Wording, rounds and deadlines change often, and an automatically scraped intermediate state can easily produce a plausible-but-wrong conclusion that you then plan two months around.

Labelling it static snapshot as of <date> hands the last judgement back to you: see something interesting, open the original source, confirm, then act.

Who it’s for

The match-strength box reads 86% because this radar was built from one specific person’s situation: experimental psychology background, master’s finishing 2027, aiming at cognitive neuroscience / BCI / Brain × AI.

If your coordinates are close to that, use it directly. If they aren’t, the structure still transfers — porting “rank by whether it produces a result within 30 days” and “every entry carries a next action plus its original source” to your own field beats assembling one more complete list.