People shaping AI work at a median of 3 organizations over a career: 3 for university starters and institute and lab starters, versus 2 for company starters
Data as of
- People with a dated work role
- 6,853
- Median organizations per career
- 3
- Median, university starters
- 3
- Median, institute and lab starters
- 3
- Median, company starters
- 2
How many organizations do people shaping AI work at over a career, how does that count vary by role group and starting sector, and how do employer counts compare across starting decades? Across 6,853 people with at least one dated work role (rolling sub-organizations up to their top-level parent organization and leaving out degrees), we estimate a career median of 3 parent organizations (middle half: 2 to 5; mean: 3.6). By role group, researchers, lab leaders and public voices have the highest median at 3 organizations (mean 3.7 for researchers, mean 3.3 for lab leaders and mean 2.9 for public voices), and builders have the lowest at 2 (mean 2.5). By starting sector, university starters and institute and lab starters worked at a median of 3 organizations, compared with 2 for company starters. Career-to-date totals peak among 1990s starters at a mean of 4.5 organizations and stand at 1.7 for 2020–2026 starters; over a fixed 5-year window from career start, the mean number of organizations was 1.5 in the 1950s and 2.1 for 2020–2021 starters.
Key findings
- Across 6,853 people with a dated work position (91% of the 7,547 individual profiles we track), 5,574 (81%) have worked at 2 or more parent organizations, 2,936 (43%) at 4 or more, and 1,160 (17%) at 6 or more, while 1,279 (19%) have worked at a single parent organization.
- By role group, researchers work at a median of 3 parent organizations (middle half 2 to 5, mean 3.7, 5,803 people), lab leaders work at a median of 3 parent organizations (middle half 2 to 4, mean 3.3, 290 people), public voices work at a median of 3 parent organizations (middle half 2 to 4, mean 2.9, 378 people) and builders work at a median of 2 parent organizations (middle half 1 to 3, mean 2.5, 382 people).
- By earliest work sector, university starters (4,282 people, mean 3.9) and institute and lab starters (927 people, mean 3.7) work at a median of 3 parent organizations, versus 2 for company starters (1,644 people, mean 2.7). Single-organization careers make up 32% of company starters (528 of 1,644), compared with 18% of institute and lab starters (170 of 927) and 14% of university starters (581 of 4,282).
- People who crossed between universities, companies, and institutes or labs (3,955 people, or 58%) worked at a median of 4 parent organizations (middle half 3 to 6, mean 4.6) — a median of 4 for university starters (2,650 people), 4 for institute and lab starters (650) and 4 for company starters (655) — compared with a median of 2 (middle half 1 to 3, mean 2.1) for the 2,898 people who stayed within their starting sector.
- Within a person's first 10 career years (among 5,388 people who started between 1950 and 2016), the mean number of parent organizations was 2.1 in the 1950s (1.8 in the 1960s), peaked at 2.8 in the 2000s, and was 2.7 in 2010–2016, while the share working at 2 or more organizations within 10 years was 68% in the 1950s and 75% in 2010–2016.
- Among the 41 organizations where at least 30 people held their earliest dated work role, Chinese Academy of Sciences leads at a mean of 5.9 career organizations (median 4.5), while NVIDIA has the lowest mean at 1.4 (median 1, with 69% staying at that single parent organization).
Employers over a career by role group and starting sector
Across 6,853 placed people, 1,279 (19%) worked at one parent organization, 1,349 (20%) worked at two parent organizations, 1,289 (19%) worked at three parent organizations, 1,076 (16%) worked at four parent organizations and 700 (10%) worked at five parent organizations, while 1,160 (17%) worked at 6 or more (the threshold for our Wayfarer badge).
Range plot showing median and interquartile parent organization counts for 6,853 people across role groups (researchers median 3, middle half 2 to 5, lab leaders median 3, middle half 2 to 4, public voices median 3, middle half 2 to 4 and builders median 2, middle half 1 to 3) and starting sectors (university starters median 3, middle half 2 to 5, institute and lab starters median 3, middle half 2 to 5 and company starters median 2, middle half 1 to 4).
Source: People of Superintelligence career records from biographies and OpenAlex, as of 22 October 2026. Sub-organizations roll up to their top-level parent organization; degrees are left out.
Data table
| Dimension | Group | People | 25th percentile | Median orgs | 75th percentile | Mean orgs | 2+ orgs | 4+ orgs | 6+ orgs |
|---|---|---|---|---|---|---|---|---|---|
| Role group | Researchers | 5,803 | 2 | 3 | 5 | 3.7 | 4,806 (83%) | 2,634 (45%) | 1,069 (18%) |
| Role group | Lab leaders | 290 | 2 | 3 | 4 | 3.3 | 230 (79%) | 105 (36%) | 42 (14%) |
| Role group | Public voices | 378 | 2 | 3 | 4 | 2.9 | 290 (77%) | 120 (32%) | 30 (8%) |
| Role group | Builders | 382 | 1 | 2 | 3 | 2.5 | 248 (65%) | 77 (20%) | 19 (5%) |
| Starting sector | Started at a university | 4,282 | 2 | 3 | 5 | 3.9 | 3,701 (86%) | 2,107 (49%) | 865 (20%) |
| Starting sector | Started at an institute or lab | 927 | 2 | 3 | 5 | 3.7 | 757 (82%) | 413 (45%) | 182 (20%) |
| Starting sector | Started at a company | 1,644 | 1 | 2 | 4 | 2.7 | 1,116 (68%) | 416 (25%) | 113 (7%) |
| Starting organization type | ↳ Government research centers | 407 | 2 | 4 | 6 | 4.3 | 356 (87%) | 223 (55%) | 103 (25%) |
| Starting organization type | ↳ Standalone research labs | 175 | 2 | 3 | 5 | 4.0 | 152 (87%) | 85 (49%) | 39 (22%) |
| Starting organization type | ↳ Nonprofits and institutes | 345 | 1 | 2 | 4 | 3.0 | 249 (72%) | 105 (30%) | 40 (12%) |
| Sector mobility | Crossed sectors over career | 3,955 | 3 | 4 | 6 | 4.6 | 3,955 (100%) | 2,501 (63%) | 1,086 (27%) |
| Sector mobility | Stayed in starting sector | 2,898 | 1 | 2 | 3 | 2.1 | 1,619 (56%) | 435 (15%) | 74 (3%) |
| All placed people | All placed people | 6,853 | 2 | 3 | 5 | 3.6 | 5,574 (81%) | 2,936 (43%) | 1,160 (17%) |
Across primary profile categories, researchers (5,803 people, or 85% of placed profiles) have a median of 3 parent organizations (middle half 2 to 5, mean 3.7), with 83% (4,806) working at 2 or more and 45% (2,634) at 4 or more. Lab leaders (290 people) and public voices (378 people) have medians of 3 and 3 organizations (means 3.3 and 2.9, with 36% and 32% working at 4 or more). Builders (382 people) have a median of 2 organizations (middle half 1 to 3, mean 2.5), with 35% (134) working at a single parent organization and 20% (77) working at 4 or more. Over a common 10-year window from career start, researchers average 2.6 organizations (76% at 2 or more), compared with 2.5 for lab leaders, 2.1 for builders and 2.1 for public voices.
Splitting the 6,853 placed people by the sector of their earliest dated work role places 4,282 (62%) at universities, 1,644 (24%) at companies and 927 (14%) at institutes and labs. University starters (median 3, middle half 2 to 5, mean 3.9) and institute and lab starters (median 3, middle half 2 to 5, mean 3.7) work at a higher median number of parent organizations than company starters (median 2, middle half 1 to 4, mean 2.7). Inside institutes and labs, people who began at government research centers (407 people, median 4, mean 4.3), standalone research labs (175 people, median 3, mean 4.0) and nonprofits and institutes (345 people, median 2, mean 3.0) differ by 1.3 organizations on the mean.
Within every role group, people who started at a university have a higher median employer count than those who started at a company: 3 versus 2 among researchers (3.9 vs. 2.8 on the mean), 4 versus 2 among lab leaders (4.7 vs. 2.5 on the mean), 3 versus 2 among public voices (3.5 vs. 2.8 on the mean) and 3 versus 2 among builders (3.5 vs. 2.2 on the mean). Across all role groups, 70% of institute and lab starters (650 of 927) and 62% of university starters (2,650 of 4,282) later worked in another sector, compared with 40% of company starters (655 of 1,644). People who crossed sectors have a median of 4 for university and for institute and lab starters and 4 for company starters (means 4.8, 4.6 and 3.9), while people who stayed in a single sector have a median of 2 organizations if they stayed in universities (1,632 people, mean 2.4) and 1 if they stayed in companies (989 people, mean 1.8) or in institutes and labs (277 people, mean 1.6).
How employer counts changed across starting decades
Grouping the 6,853 placed people by the decade of their earliest dated work role yields eight cohorts with at least 50 people, covering 6,822 people from the 1950s (75 people) through 2020–2026 (872 people). Another 31 people began working before 1950. We compare full career-to-date totals alongside fixed 5-year and 10-year windows measured from each person's career start.
Line chart across 8 starting decades (1950s to 2020–2026) showing mean parent organizations worked at over a career to date and within the first 5 years of a career.
Source: People of Superintelligence career records from biographies and OpenAlex, as of 22 October 2026. The 5-year window counts organizations started in the first 5 calendar years of a career among people who began working in or before 2021.
Data table
| Starting decade | People | Career median (mean) | Career 2+ orgs | Career 4+ orgs | First 5y mean (2+ share) | First 10y mean (2+ share) |
|---|---|---|---|---|---|---|
| 1950s | 75 | 3 (3.6) | 69 (92%) | 35 (47%) | 1.5 (40%, n=75) | 2.1 (68%, n=75) |
| 1960s | 210 | 4 (3.8) | 186 (89%) | 109 (52%) | 1.5 (37%, n=210) | 1.8 (57%, n=210) |
| 1970s | 451 | 4 (4.1) | 410 (91%) | 237 (53%) | 1.7 (49%, n=451) | 2.2 (68%, n=451) |
| 1980s | 826 | 4 (4.2) | 762 (92%) | 458 (55%) | 1.7 (52%, n=826) | 2.2 (71%, n=826) |
| 1990s | 1,204 | 4 (4.5) | 1,110 (92%) | 709 (59%) | 1.9 (59%, n=1,204) | 2.5 (78%, n=1,204) |
| 2000s | 1,467 | 4 (4.2) | 1,321 (90%) | 832 (57%) | 2.0 (63%, n=1,467) | 2.8 (80%, n=1,467) |
| 2010s | 1,717 | 3 (2.8) | 1,321 (77%) | 481 (28%) | 1.9 (57%, n=1,717) | 2.7 (75%, n=1,155) |
| 2020–2026 | 872 | 1 (1.7) | 369 (42%) | 60 (7%) | 2.1 (60%, n=290) | — |
In full career-to-date counts, the median number of parent organizations is 3 for 1950s starters (mean 3.6), 4 for 1990s starters — where the career-to-date mean peaks at 4.5 organizations (59% working at 4 or more) — 3 (mean 2.8, 28% at 4 or more) for 2010s starters, and 1 (mean 1.7, 7% at 4 or more) for 2020–2026 starters.
Over an equal 5-year span from career start (covering people who began working in or before 2021), the 5-year median is 1 organization in the 1950s and 2 in 2020–2021. The 5-year mean was 1.5 in the 1950s and the 1960s, 1.9 in the 2010s, and 2.1 in 2020–2021. The share working at 2 or more organizations in their first 5 years was 40% (30 of 75) in the 1950s and 60% (174 of 290) in 2020–2021, and the share working at 3 or more within 5 years was 7% (5 of 75) in the 1950s and 32% (94 of 290) in 2020–2021.
Over a 10-year window from career start (covering 5,388 people who began working between 1950 and 2016), the median number of parent organizations was 2 in the 1950s and 2 in 2010–2016. The 10-year mean was 2.1 in the 1950s (1.8 in the 1960s), peaked at 2.8 in the 2000s (80% working at 2 or more, 52% at 3 or more), and was 2.7 in 2010–2016 (75% at 2 or more, 49% at 3 or more). By starting sector, the 10-year mean in the 1950s versus 2010–2016 was 2.0 versus 2.8 for university starters, 2.1 versus 2.9 for institute and lab starters, and 2.1 versus 2.3 for company starters.
Where multi-employer and single-employer careers began
Across the 41 parent organizations where at least 30 people began their dated careers, Massachusetts Institute of Technology (285 starters) has the largest starting cohort.
Horizontal bar chart of 41 starting organizations with at least 30 career starters, ranked by mean parent organizations worked at over a career.
Source: People of Superintelligence career records from biographies and OpenAlex, as of 22 October 2026. A person with simultaneous first-year roles at two parent organizations counts in both.
Data table
| Starting organization | Sector | Career starters | Median start year | Median orgs (IQR) | Mean orgs | Stayed at 1 org | 4+ orgs | Crossed sectors | First 10y mean |
|---|---|---|---|---|---|---|---|---|---|
| Chinese Academy of Sciences | Institute or lab | 32 | 2005 | 4.5 (3–8) | 5.9 | 2 (6%) | 19 (59%) | 29 (91%) | 3.5 (n=29) |
| CNRS | Institute or lab | 43 | 1995 | 5 (3–6.5) | 5.3 | 1 (2%) | 30 (70%) | 40 (93%) | 3.5 (n=40) |
| University of Tokyo | University | 30 | 2004 | 4.5 (4–5) | 4.9 | 0 (0%) | 24 (80%) | 27 (90%) | 3.0 (n=28) |
| Tsinghua University | University | 38 | 2004 | 4 (3–6) | 4.8 | 4 (11%) | 22 (58%) | 28 (74%) | 2.9 (n=29) |
| Inria | Institute or lab | 46 | 2003 | 4 (3–6) | 4.6 | 6 (13%) | 31 (67%) | 39 (85%) | 3.4 (n=38) |
| Max Planck Society | Institute or lab | 46 | 2006 | 4 (3–6) | 4.6 | 4 (9%) | 29 (63%) | 41 (89%) | 3.3 (n=39) |
| University of Illinois Urbana-Champaign | University | 62 | 1997 | 4 (3–5.8) | 4.3 | 5 (8%) | 37 (60%) | 36 (58%) | 2.8 (n=58) |
| University of Maryland | University | 39 | 2003 | 4 (3–5.5) | 4.2 | 5 (13%) | 22 (56%) | 24 (62%) | 3.0 (n=36) |
| University of Edinburgh | University | 55 | 1990 | 4 (2–5) | 4.1 | 4 (7%) | 29 (53%) | 35 (64%) | 2.6 (n=53) |
| California Institute of Technology | University | 42 | 1994 | 4 (2–5.8) | 3.9 | 7 (17%) | 25 (60%) | 29 (69%) | 2.8 (n=38) |
| University of California, Los Angeles | University | 36 | 1997 | 3.5 (3–5) | 3.9 | 3 (8%) | 18 (50%) | 27 (75%) | 2.4 (n=35) |
| University of Pennsylvania | University | 57 | 1992 | 3 (3–5) | 3.9 | 6 (11%) | 28 (49%) | 36 (63%) | 2.5 (n=54) |
| University of Massachusetts Amherst | University | 30 | 1991 | 3 (2–4) | 3.9 | 1 (3%) | 13 (43%) | 18 (60%) | 2.4 (n=29) |
| Columbia University | University | 45 | 2006 | 4 (2–5) | 3.8 | 5 (11%) | 26 (58%) | 29 (64%) | 2.8 (n=37) |
| Princeton University | University | 60 | 2005 | 4 (2.8–5) | 3.7 | 7 (12%) | 32 (53%) | 31 (52%) | 2.9 (n=48) |
| University of Toronto | University | 72 | 2008 | 3 (2.8–5) | 3.7 | 9 (13%) | 31 (43%) | 58 (81%) | 2.9 (n=54) |
| Yale University | University | 39 | 1987 | 3 (2–5) | 3.7 | 3 (8%) | 19 (49%) | 23 (59%) | 2.4 (n=36) |
| ETH Zurich | University | 38 | 2011 | 4 (2.3–4) | 3.6 | 3 (8%) | 20 (53%) | 22 (58%) | 2.9 (n=29) |
| Massachusetts Institute of Technology | University | 285 | 1993 | 3 (2–5) | 3.6 | 25 (9%) | 132 (46%) | 188 (66%) | 2.5 (n=260) |
| Stanford University | University | 178 | 2001 | 3 (2–4) | 3.6 | 13 (7%) | 82 (46%) | 116 (65%) | 2.7 (n=153) |
| University of California, Berkeley | University | 165 | 2009 | 3 (2–5) | 3.6 | 18 (11%) | 74 (45%) | 112 (68%) | 2.7 (n=125) |
| University of Cambridge | University | 68 | 1999 | 3 (2–4.3) | 3.6 | 7 (10%) | 26 (38%) | 50 (74%) | 2.6 (n=55) |
| University of Southern California | University | 38 | 1991 | 3 (2–4) | 3.6 | 5 (13%) | 18 (47%) | 26 (68%) | 2.4 (n=36) |
| Georgia Institute of Technology | University | 38 | 2011 | 4 (1.3–5) | 3.5 | 10 (26%) | 20 (53%) | 17 (45%) | 2.4 (n=28) |
| Carnegie Mellon University | University | 159 | 1996 | 3 (2–5) | 3.5 | 22 (14%) | 68 (43%) | 102 (64%) | 2.4 (n=142) |
| Harvard University | University | 77 | 2004 | 3 (2–4) | 3.5 | 10 (13%) | 35 (45%) | 45 (58%) | 2.5 (n=67) |
| AT&T | Company | 51 | 1987 | 3 (2.5–4) | 3.5 | 0 (0%) | 24 (47%) | 43 (84%) | 2.2 (n=51) |
| University of Texas at Austin | University | 38 | 2000 | 3 (2–4.8) | 3.5 | 6 (16%) | 18 (47%) | 21 (55%) | 2.4 (n=32) |
| New York University | University | 34 | 2008 | 3 (2–4) | 3.5 | 5 (15%) | 15 (44%) | 22 (65%) | 2.6 (n=25) |
| University of Oxford | University | 85 | 2007 | 3 (2–4) | 3.4 | 15 (18%) | 37 (44%) | 55 (65%) | 2.6 (n=63) |
| IBM | Company | 83 | 1994 | 3 (2–4) | 3.4 | 19 (23%) | 31 (37%) | 46 (55%) | 2.1 (n=74) |
| National University of Singapore | University | 31 | 2008 | 3 (1.5–4.5) | 3.4 | 8 (26%) | 13 (42%) | 14 (45%) | 2.8 (n=24) |
| Cornell University | University | 64 | 2002 | 3 (1–4) | 3.2 | 17 (27%) | 25 (39%) | 34 (53%) | 2.7 (n=47) |
| EPFL | University | 32 | 2008 | 3 (1.8–4) | 3.2 | 8 (25%) | 14 (44%) | 16 (50%) | 2.7 (n=27) |
| Microsoft | Company | 116 | 2015 | 2 (1–4) | 2.9 | 32 (28%) | 35 (30%) | 58 (50%) | 2.5 (n=66) |
| University of Washington | University | 55 | 2009 | 2 (1.5–4) | 2.8 | 14 (25%) | 18 (33%) | 33 (60%) | 2.0 (n=37) |
| Meta | Company | 85 | 2018 | 2 (1–3) | 2.1 | 30 (35%) | 13 (15%) | 22 (26%) | 2.2 (n=23) |
| Company | 183 | 2017 | 2 (1–2) | 1.9 | 89 (49%) | 15 (8%) | 41 (22%) | 1.9 (n=84) | |
| OpenAI | Company | 44 | 2020 | 1 (1–2) | 1.7 | 25 (57%) | 3 (7%) | 4 (9%) | 2.6 (n=9) |
| Allen Institute for AI (Ai2) | Institute or lab | 33 | 2022 | 2 (1–2) | 1.6 | 16 (48%) | 1 (3%) | 17 (52%) | 2.0 (n=2) |
| NVIDIA | Company | 32 | 2022 | 1 (1–2) | 1.4 | 22 (69%) | 0 (0%) | 5 (16%) | 1.1 (n=7) |
Research institutes with at least 30 starters (Chinese Academy of Sciences at a mean of 5.9 organizations, median 4.5, 32 starters, 91% crossing sectors, CNRS at a mean of 5.3 organizations, median 5, 43 starters, 93% crossing sectors, Inria at a mean of 4.6 organizations, median 4, 46 starters, 85% crossing sectors, Max Planck Society at a mean of 4.6 organizations, median 4, 46 starters, 89% crossing sectors and Allen Institute for AI (Ai2) at a mean of 1.6 organizations, median 2, 33 starters, 52% crossing sectors) are ranked 1st, 2nd, 5th, 6th and 40th of 41. Among the 29 universities with at least 30 starters, mean career organization counts run from 2.8 at University of Washington (55 starters, median 2) to 4.9 at University of Tokyo (30 starters, median 4.5).
Among the 7 companies with at least 30 career starters, mean career organization counts are 3.5 at AT&T (51 starters, median start year 1987, median 3 organizations), 3.4 at IBM (83 starters, median start year 1994, median 3 organizations), 2.9 at Microsoft (116 starters, median start year 2015, median 2 organizations), 2.1 at Meta (85 starters, median start year 2018, median 2 organizations), 1.9 at Google (183 starters, median start year 2017, median 2 organizations), 1.7 at OpenAI (44 starters, median start year 2020, median 1 organization) and 1.4 at NVIDIA (32 starters, median start year 2022, median 1 organization). NVIDIA has the highest single-employer retention rate among all 41 leading starting organizations, with 69% (22 of 32) of its starters holding no dated work role at any other parent organization.
The highest dated biography organization counts in each role group belong to Klaus Peter Jantke (13 dated biography organizations), followed by Anatole von Lilienfeld (12) among researchers; Olaf Groth (10 dated biography organizations), followed by Ann O'Leary (8) and Eric Schmidt (8) among public voices; Bruce Wilcox (8 dated biography organizations), followed by Jeff Black (7), John Krafcik (7), Kriti Sharma (7), Kyle Vogt (7) and Paul Kruszewski (7) among builders; and Louis Rosenberg (7 dated biography organizations), followed by Adebayo Ogunlesi (6), Bowen Zhou (6), Emmett Shear (6), Kristin E. Lauter (6), Mérouane Debbah (6), Peter Jackson (6), Ren Ito (6), Tareq Amin (6), Yinfei Yang (6) and Zoubin Ghahramani (6) among lab leaders. Among single-organization careers, 114 people currently hold a position at the single parent organization where they have worked for 20 or more recorded years with no other employer in their career file, led by Thomas B. Sheridan at Massachusetts Institute of Technology (68 years, since 1959), Carver Mead at California Institute of Technology (67 years, since 1960), Ronald Fagin at IBM (54 years, since 1973), Yuen Poovarawan at Kasetsart University (54 years, since 1973) and Zygmunt Vetulani at Adam Mickiewicz University (54 years, since 1973).
Career data comes from the biographies on this site and from OpenAlex author records, as of 22 October 2026. People whose OpenAlex record may be incomplete or mixed are left out of citation figures.