Skip to content
DeeptechNavigator

Data coverage

What is in the platform, layer by layer

Public registries give names, titles and dates. Each layer below is work we added on top of them: reading, judging, matching, profiling and mapping capital. Each shows how much we hold, how full it is, and the caveat that goes with it. Counts and shares only; the records themselves are in the plans.

Data as of 12 Oct 2026Methodology

Totals rebuilt weekly from our records, last on 12 Oct 2026. How we build them →

Six layers, one dataset

The bottom layer is the public record. The five above it are ours.

  1. 0 · Collected4,84,093Indian patent applications in our set84% filed from 2022 onward. Indian Patent Office records only. Designs run from the first year shown; trademarks are held for the applicants we match against, not the whole register.
  2. 1 · Read4,62,986patent applications read by AI and placed in a sector, of 4,84,093 in our set96% of the set readAI-generated summary, not the patent text - a starting point; verify before relying.
  3. 2 · Judged29,513startups judged deep tech, confirmed or likely, of 4,53,422 registered startups in our set7% of registered startups in our setRead by AI, then checked by our team: we review samples and decide the hard cases by hand. Each company profile shows the evidence its verdict rests on.
  4. 3 · Matched8,936deep-tech startups matched to a patent, design or trademarkA patent application is matched for 5,000 deep-tech startups.30% of deep-tech startupsStartup counts are name-matched candidates, not ownership records - see methodology.
  5. 4 · Profiled87%of deep-tech startups in our set have a web profile read by AI and matchedRead by AI from the company's own site and public sources, then checked by our team through samples and with our users. A record can still be incomplete or matched to the wrong company; check a figure before relying on it.
  6. 5 · Mapped2,037funding rounds on record, at 1,650 companiesAcross all companies in our set, not only deep tech.Funding is from public sources and may contain errors - verify before relying on it.

0 · Collected

Public record: what the registries give

Public record: Patent, design and trademark journals from the Indian Patent Office, and the startup registry: titles, applicants, dates and status; names, states and sectors.

We added: This is the base the layers above are built on.

of the patent applications granted
1,11,613
of the patent applications granted
industrial designs in our set, filed from 2024
76,820
industrial designs in our set, filed from 2024
trademark filings in our set
2,86,064
trademark filings in our set

Patent applications placed in a sector, by filing year

Indian applicantForeign applicant

Patent applications placed in a sector, by filing year
YearIndian applicantForeign applicant
202226,40730,660
202344,49132,413
202458,67635,537
202567,41833,317
2026 (so far)48,31710,820

Chart starts at 2022 - earlier years reflect granted-only survivorship, not full filing volume.

The latest year is partial and still filling in: patents publish about 18 months after filing, and rounds are added as they are reported.

Indian Patent Office records only. Designs run from the first year shown; trademarks are held for the applicants we match against, not the whole register.

Deeptech Navigator · Data coverage · rebuilt 12 Oct 2026

1 · Read

Read: each patent placed in a sector and a technology

Public record: A title, an abstract, an applicant and a status.

We added: The problem each one solves and its approach, in 225 technologies across 14 sectors.

Patent applications by sector

All filers in our set: Indian and foreign, companies and institutes. Other holds filings not yet placed in one of the named sectors.

Patent applications by sector
LabelPatent applications
Health & Life Sciences
1,07,61623%
Communications
56,84112%
Mobility & Transport
50,13311%
Computing & AI
48,20410%
Energy
47,80810%
Agri & Food
30,4757%
Materials & Nanotech
30,0356%
Robotics & Industrial Automation
24,3705%
Climate & Environment
17,6854%
Fintech & Consumer
17,4114%
Defence & Security
14,1673%
Other
10,6302%
Semiconductors
6,1221%
Space
9640%
Quantum
5250%

Technology trends by sector →

AI-generated summary, not the patent text - a starting point; verify before relying.

Deeptech Navigator · Data coverage · rebuilt 12 Oct 2026

2 · Judged

Judged: which registered startups are deep tech

Public record: A name, a state and a sector the company picked.

We added: A verdict that names its evidence, from the company's own site to the registry text alone.

Deep-tech startups by sector

Deep-tech startups by sector
LabelDeep-tech startups
Health & Life Sciences
5,77120%
Communications
2921%
Mobility & Transport
3,23511%
Computing & AI
3,07310%
Energy
3,30211%
Agri & Food
1,4005%
Materials & Nanotech
2,87610%
Robotics & Industrial Automation
4,74416%
Climate & Environment
6312%
Fintech & Consumer
3101%
Defence & Security
6872%
Other
1,9146%
Semiconductors
4502%
Space
6212%
Quantum
2071%

Deep-tech startups by state

Leading states

  1. 1Maharashtra5,024
  2. 2Karnataka4,353
  3. 3Tamil Nadu3,211
  4. 4Telangana2,388
  5. 5Delhi2,258
  6. 6Uttar Pradesh2,226
FewerMore

1 - 5,024 deep-tech startups per state

14 deep-tech startups in locations not shown on the map (union territories or unrecognised state labels).

How a verdict is reached →

Read by AI, then checked by our team: we review samples and decide the hard cases by hand. Each company profile shows the evidence its verdict rests on.

Deeptech Navigator · Data coverage · rebuilt 12 Oct 2026

3 · Matched

Matched: the startup behind a filing

Public record: An applicant's name on a filing. Nothing says which startup it is.

We added: The registered startup behind the name, with a confidence label on each match.

The broader figure includes trademarks, which on their own say little about the technology; a patent match is counted apart.

Share of deep-tech startups matched to a patent, design or trademark, in the eight states with the most

Share of deep-tech startups matched to a patent, design or trademark, in the eight states with the most
Maharashtra33%
Karnataka34%
Tamil Nadu26%
Telangana29%
Delhi35%
Uttar Pradesh28%
Gujarat34%
Kerala25%

How matching works →

Startup counts are name-matched candidates, not ownership records - see methodology.

Deeptech Navigator · Data coverage · rebuilt 12 Oct 2026

4 · Profiled

Profiled: what a deep-tech profile holds

Public record: A registered address and, for some, a website.

We added: What each company makes, its brand and the customers it names: 1,25,231 web profiles read and matched across all companies in our set. Directors and incorporation year come from the company register where we matched it.

Shares of the deep-tech startups in our set, confirmed or likely and not closed. A share counts what is on record with us, not what exists, and is recounted within the hour, so it can differ slightly from the weekly totals.

Read by AI from the company's own site and public sources, then checked by our team through samples and with our users. A record can still be incomplete or matched to the wrong company; check a figure before relying on it.

Deeptech Navigator · Data coverage · profile shares cached for about an hour

5 · Mapped

Mapped: the capital around the companies

Public record: News stories on some rounds, and lists of investors.

We added: Rounds tied to the companies that raised them, and the investors, incubators and family offices around them.

Funding rounds on record, by year

Funding rounds on record, by year
YearRounds
201967
202074
2021135
2022187
2023208
2024302
2025395
2026 (so far)348

180 rounds dated before 2019 are counted in the total above but not drawn.

141 rounds carry no date and are not drawn.

investors by focus, stage and state
1,243
investors by focus, stage and state
incubators
1,390
incubators
accelerators
225
accelerators
family offices
261
family offices

11 of the family offices are backing a company we call deep tech by a confirmed match.

41 more are possible, matched on a trading brand only, and not counted as backing.

Funding by sector →Family offices (members) →

Funding is from public sources and may contain errors - verify before relying on it.

The latest year is partial and still filling in: patents publish about 18 months after filing, and rounds are added as they are reported.

Family offices are listed from Inc42's Family Office Tracker (April 2026) and from funding rounds and news reports, each with the source it came from.

Backing deep tech means a company it is known for is one we call deep tech, matched by registered name or a funding round. A match on a trading brand alone is counted apart, as possible.

Deeptech Navigator · Data coverage · rounds rebuilt 12 Oct 2026

People

Reviewed by our team

AI does the first read at this scale, and our team stands behind what it produces. We set the rules it follows, check samples of its work by hand, check findings with the investors who use them, and decide the hard cases ourselves. Each decision is kept with who made it and when.

What our team decides by hand

The deep-tech verdict on the hard cases, which fund-manager entries are one firm, and which technology names are the same.

Rules our team set, which the pipeline follows

  • A not-for-profit, a government company or a unit of a large company is not listed as a deep-tech startup.
  • A company matched to a fund on its trading brand alone is marked not confirmed, and that match is not counted as funding.
  • Campuses of one university, and the institutes of one system such as the IITs, are kept apart, never merged.
  • A company whose own website is parked or taken over by another business is not judged deep tech from it.

Samples checked by hand

From October 2026 our team checks random samples of the AI's company profiles, patent summaries and deep-tech verdicts by hand, marking each right, partly right or wrong. The results show here once 20 have been checked.

Between those checks, a review runs at two-hour intervals: it samples what the pipeline wrote and flags anything that looks wrong, and our team decides what to change.

How often each layer is updated

Public recordUpdate cycle: Monthly to quarterly, as the journals and the registry publish; figures rebuilt weekly
ReadUpdate cycle: New applications read after each import; figures rebuilt weekly
JudgedUpdate cycle: Rebuilt weekly
MatchedUpdate cycle: Rebuilt weekly
ProfiledUpdate cycle: Company websites read continuously; figures updated within the hour
MappedUpdate cycle: Funding news read continuously; rounds rebuilt 12 Oct 2026

How recent: the latest record on file

Designs published to
31 Jul 2026
Trademarks filed to
08 Jun 2026
Startups registered to
18 Jun 2026
Funding rounds to
29 Sept 2026
Company websites read to
12 Oct 2026

Sources are refreshed monthly to quarterly; company websites and funding news are read continuously.

How verdicts, matches and rankings are built →

Counts and shares here; the records are in the plans

This page shows how much we hold and how full it is. Company lists, profiles, filings and rounds are in the plans. A study on one company or one field is commissioned apart.

Commission a study →Connect AI tools →See plans and prices →Book a demo →

Access is granted manually by the team - we reply by email, typically within 1 business day.