One chart: every top-level org under Bill Dally and what sits inside it · reconstructed from public sources · 2026-07-29
org-chartnvidia-researchcareersreconstructed
1 · TL;DR
NVIDIA never publishes an org chart. This is reconstructed from lab pages, personal
pages and publication rosters. Leads are reliable. Headcounts are estimates.
Ten orgs report to Bill Dally, and the structure is flat — a new research
scientist is usually two hops from a VP.
Only LPR publishes its internal group list. Everything labelled
thrusts below is a set of research directions, not reporting units — one person
commonly spans several.
2 · The chart
Left column: the ten organizations, with lead and headcount. Right column: what sits inside each. The groups / thrusts label on each org box says whether those rows are real reporting units or just published research directions.
3 · How much of this is verified
Evidence tier
Applies to
Confidence
Named member roster on an official lab page
GenAIR (9), DAIR (13), AMRI (10), SRL (17+11), AV (23+3+15), RTR (30+6)
High — counted directly
Official thrust list, no roster
Cosmos, ADLR, GEAR, AV, SRL, RTR
Medium — directions real, boxes not org units
Inferred from publications and personal pages
All ~ headcounts; DLER and Zhiding Yu group sizes; AMRI's four themes
Low — order of magnitude only
Genuinely ambiguous
RVP's parent org — Birchfield and his people appear under both LPR and SRL
Unresolved
Two facts are hard, not inferred: /labs/dir/ (deep imagination research)
301-redirects to /labs/cosmos-lab/, confirming the rename; and DAIR's own page
states it is part of LPR, while AMRI's does not.
Orgs publishing full rosters are the ones that need visibility to recruit. AV research
(40+ alumni listed) and SRL (11 interns, schools named) are the most open to cold email.
Orgs with no roster — Cosmos, ADLR, GEAR — hire through referral and internal transfer.
4 · Where the orgs overlap
Neural rendering is contested by four orgs — SIL, AMRI, RTR and Cosmos all publish in it
and regularly compete at the same SIGGRAPH session.
This is deliberate. Dally's model runs redundant attacks on one problem from different angles
rather than assigning a single owner.
So the useful question is not "which org does 3D" but what the 3D is for: simulation →
SIL, humans → AMRI, rendering → RTR, robots → SRL and DAIR, video generation → Cosmos.
Robotics alone is split across four places — SRL, GEAR, DAIR, RVP — with no boundary visible
from outside.
Individuals cross-list constantly (Sifei Liu, Hongxu Yin, Stan Birchfield, Bowen Wen appear on
two or three lab pages). Lab pages are branding, not an org chart.
SIL grew by absorbing standalone labs — Ken Museth's physics team (Apr 2024), Laura Leal-Taixé's
team (Dec 2024) — which is why both still appear separately in NVIDIA's 21-lab listing.
5 · Open questions
Does RVP report to LPR or to SRL?
SRL lists Yashraj Narang as team lead while Dieter Fox is senior director — promotion,
delegation, or a stale page?
Where does AMRI's "new experiences" org sit — direct to Dally, or under another VP?
ADLR's headcount is the weakest number here; much of that org never appears on
research.nvidia.com at all.
Nothing here comes from anyone inside NVIDIA. Confirm any specific reporting line with someone
in the group before acting on it.
Reconstructed 2026-07-29 from research.nvidia.com lab and person pages, individual lab sites, and public reporting on NVIDIA Research headcount. No internal sources. Leads and counts marked "listed" are verifiable; every value marked ~ is an estimate.