Diligence brief · prepared for Ammar

Index Robotics (New York — indexrobotics.ai)

Friday, September 18, 2026

How to read this. Every fact below was gathered this afternoon, September 18, 2026, from public sources — the companies' own sites, arXiv, GitHub, and the paper project pages, all listed at the end. Nothing here comes from the company, from a data room, or from anyone who works there. Anything that is inference rather than a read fact is marked read / inference, and what is simply not public is listed under Explicitly unverified.

Bottom line

1. The headline finding: two companies share this name

This is the first thing to get straight, because the public record has them tangled.

indexrobotics.ai — New York This is the one you are looking at

Confirmed on your call: "the company founded by Raunaq."

  • Homepage is a single line: "Adaptive robotic systems to automate precise tasks." read from the site
  • Contact: info@indexrobots.ai. They run two domains — indexrobotics.ai and indexrobots.ai serve the identical page, and the CEO's email is on the second. Worth knowing so you don't read a mismatch as a red flag. both fetched today
  • No product page, no customers, no pricing, no press.
  • Still stealth: co-founder Siddhant Haldar's LinkedIn reads "Co-founder/CTO at a stealth robotics startup."
  • Age: their GitHub organisation github.com/index-robotics was created November 18, 2025 — so roughly ten months old. Bhirangi's LinkedIn shows a January 2026 start; the org predates it, which is why the ten-month figure is the one to use. GitHub API, read today

index-robotics.com — Seongnam, Bundang-gu, South Korea A different company

  • Bilingual site. Tagline: "INtelligent & DEXterous / Precision hardware for Physical AI / We build the body for intelligence."
  • Products: IDX-Handy, actuators, grippers & hands, a data glove. Their own site says IDX-Handy is "Unveiling Soon — 2H 2026", i.e. not shipped.
  • Contact sh.shin@index-robotics.com. Founder Seunghoon Shin, ex-KAIST Hubo Lab. Founded 2023.

Why it matters

The $1.0M seed (February 2026, led by FuturePlay, with 500 Global / Base / Bass Ventures; reported as 1.5 billion won in Korean press) belongs to the Korean company. Crunchbase, PitchBook and Dealroom appear to have blended the two records — including a New York HQ sitting on a Korean company's profile.

Treat any Index Robotics funding number you find online as the Korean company's until proven otherwise. There is no published round for the New York company at all.

2. The New York team — seven people, four co-founders

All seven are named on the company's own team page. read from indexrobotics.ai

inference This is a shipping-and-selling team, not a lab. A founding commercial hire at seven people means they intend to be selling something now — worth asking what.

3. What their public code says they are actually building

The company's GitHub org, github.com/index-robotics, has four public repositories. read from the GitHub API today

RepoWhat it isCreated
ml-interviewHiring loop — ML interview templateJan 6, 2026
onsite-robotics-interviewHiring loop — onsite robotics + ML templateFeb 16, 2026
deoxys_control"Internal copy of the deoxys_control library" — the Franka arm control stackJan 20, 2026
pick-place-challenge"Open-ended Franka + Robotiq pick-and-place sandbox on mjlab… For candidates to hack on."Jun 8, 2026

Franka arm. Robotiq gripper. MuJoCo / mjlab simulation. inference That is the vocabulary of precise industrial arm manipulation and sim-to-real — not humanoids and not hands. Two of the four repos are explicitly hiring artifacts and a third is an internal copy of someone else's library, so this is a read of the stack they work in, not of a product. But it is the only engineering surface they have exposed, and it points somewhere more grounded and more sellable than the humanoid-hand story the founders' research pedigree suggests.

And the CEO's own site says the same thing. Raunaq Bhirangi's personal page (raunaqbhirangi.github.io, read today) describes Index as "building adaptive systems to automate contact-rich, industrial work." read today That is the clearest public statement of the business anywhere — clearer than the company homepage — and it lines up with the Franka/Robotiq/mjlab stack rather than with humanoid hands. The same page puts his own email at raunaq@indexrobots.ai.

The gap is still worth naming on your call: the research lineage is tactile sensing and hands; the company's visible engineering is arms doing precise pick-and-place. Ask which one the company is.

What is not on their GitHub: no product code, no policies, no data. Which is the answer to "what's proprietary?" — everything that matters is off GitHub, and therefore only visible to you in diligence.

4. What this team has published — all open, all deliberately cheap

Attribution nuance worth knowing: the original RUKA paper's author list is the NYU lab (Zorin, Guzey, Yan, Iyer, Kondrich, Bhattasali, Pinto) and does not include the Index founders; Bhirangi is an author on RUKA-v2. So "they built RUKA" is true of v2 and of the lab, not of v1's byline. read from the papers today

Maintenance, measured: last pushes as read today — eFlesh Oct 28, 2025; RUKA-v2 Apr 19, 2026; anyskin Nov 2024; reskin_sensor Jun 2023. These are living reference designs with real adoption, but the commits have moved elsewhere — consistent with the team's energy now being inside a private repo.

5. The bull case

They built the default open tactile stack. The field standardized on their sensor designs before they sold anything — that is real distribution and real credibility in exactly the layer everyone needs next.

And they are positioned under the robotic-foundation-model companies rather than against them. Robotic foundation models took roughly $3.92B across nine deals this year, with Physical Intelligence alone around $1.1B raised at roughly $5.6B. The policy layer is drowning in capital; the cheap touch-and-hands layer underneath it is the less-crowded seat.

If the GitHub read is right, the arm-manipulation angle strengthens this: precise industrial pick-and-place has buyers with budgets today, which humanoid hands do not.

6. The bear case

7. The questions to ask them

  1. You've open-sourced the hand and three sensors — what are you selling?
  2. Is there a named customer or a signed pilot, or is it still demos? Their own framing is about the gap between demo and deployment — that is the right fight, and the one the whole field is losing.
  3. What's proprietary — the data corpus, the policies, or a cost advantage in manufacturing?
  4. Are you a hands-and-touch company or an industrial arm-manipulation company? Your public repos are Franka + Robotiq + mjlab; your papers are tactile skins and humanoid hands. Which is the business?
  5. What have you actually raised, and from whom? Nothing is public, so the terms are the only source of truth. (And confirm out loud that the reported $1.0M seed is not yours — it belongs to the Korean company of the same name.)

8. Verdict

At a seed price, with that team, this is worth taking seriously — conditional on a named customer or a signed pilot. Without one, it is a great-team-pre-everything bet and should be priced as such.

9. Explicitly unverified

None of the following is public, and none of it is asserted anywhere above:

Sources

All read on September 18, 2026.