Our Company
Built in India. Designed for Sovereign Nations.
Tosh Defence is an India-born sovereign defence AI and mission software company building the Sovereign AI Operating Layer for Defence, National Security and Critical Systems. It connects data, models, agents, communications and mission applications inside the customer's own security boundary, for on-premises, air-gapped and disconnected environments where external-cloud dependence is unacceptable.
OUR MISSION
Our Mission
Build sovereign AI and mission software that gives defence, national-security, government and critical-system operators trusted intelligence and operational control inside their own security boundary.
Our Vision
Become the defining provider of the Sovereign AI Operating Layer for Defence, National Security and Critical Systems, serving nations that require advanced AI without strategic dependence on foreign clouds, closed foreign software stacks or externally controlled operational infrastructure.
Our Promise
You retain control of the infrastructure, data, access, models, policies and consequential decisions.
WHO WE SERVE
Built for High-Trust Environments, Worldwide
Our first market is India. Our destination is every allied and friendly sovereign nation that needs customer-controlled defence AI and mission software.
Armed Forces
Military commands and support organisations that need mission software they control, including in disconnected and contested conditions.
Intelligence and National Security
Agencies that need compartmentation, provenance, human authority and minimal external dependency.
Police, Border and Coast Guard
Police, paramilitary, border and maritime organisations coordinating sensitive operations across dispersed units.
Government and Public Sector
Ministries and public bodies that need jurisdictional clarity, data control, audit and long-term continuity, in India and in friendly sovereign nations.
Critical Infrastructure
Energy, telecommunications, transport, finance and public-service operators that need resilience, visibility and local control.
Regulated Enterprises
Organisations that need customer-controlled AI, records intelligence and compliance evidence inside their own boundary.
OUR VALUES
What We Stand For
Built in India, Designed for Sovereign Nations
India is our origin and first market, not our ceiling. We build for any nation that needs advanced AI without strategic dependence on external control.
Customer Control by Default
Infrastructure, data, access, models, policies and consequential decisions stay with you. Data residency alone is not enough.
Evidence Over Adjectives
We describe architecture, test results and limitations rather than superlatives, and we keep current capability, roadmap and concept clearly separate.
Defensive and Human-Authorised
Defensive, dual-use and decision-support software only. People, not agents, authorise consequential action.
Operator-Centric
We design for the operators who use these systems under real constraints: degraded networks, limited hardware and high stakes.
Lawful and Responsible
International work follows export controls, destination-country rules, end-use checks and partner due diligence.
COMPETITIVE EDGE
Why Tosh Defence
We do not need to win the frontier-model race. We build the controlled system that makes AI usable inside a sovereign boundary.
We Compete Above the Model Layer
Models are replaceable components. The durable product is the controlled system around them: deployment, data, retrieval with citations, orchestration, approval and audit.
Demonstrable Customer Control
A deployment and assurance architecture designed so the operator can show who controls each layer.
Reusable Data Models and Adapters
Canonical data models and adapter contracts that carry across mission domains and hardware vendors.
Governed AI and Agentic Runtime
Allow-listed tools, visible plans, human approval and attributable audit for every agent action.
Many Applications, One Footprint
Once one application is accredited, the next extends a footprint you already trust.
Problem Statement to Bounded System
We convert a stated operational need into a scoped, working and measurable system, with its limits written down.
OUR JOURNEY
Key Milestones
Founded
Tosh Defence Private Limited is established in Gurugram, India, to build sovereign AI and mission software for defence and national security.
SANKET Development Begins
Work starts on SANKET, a sovereign secure information exchange platform built on Signal Protocol through official libsignal bindings.
Portfolio Expands
MAYA cyber deception, DRISHYA infrastructure monitoring and SAJAG predictive maintenance enter development alongside RecordsKeeper.AI and Tosh AI.
The Sovereign AI Operating Layer
The six mission applications are brought together on one operating layer for deployment, data, intelligence, orchestration and assurance, for India first and for sovereign nations worldwide.
Indigenising the Software Mind
India has worked hard to indigenise the metal of defence while still importing too much of its software mind. Tosh Defence exists to change that. It is our origin story, not our geographical ceiling: the same problem exists wherever a nation cannot accept that its sensitive data, models and decision workflows depend on an external vendor or jurisdiction.
LEADERSHIP
Our Team
Rohendra Singh
Programmer, architect and DevOps engineer with 15 years of experience building software products, including more than seven secure SaaS applications.
Kanishka Sharma
A results-driven Operations Head with expertise in customer success, client experience optimisation and retention. Skilled in problem-solving, communication and relationship-building.
Sharjeel Maniyar
A product designer with experience in interaction design and front-end development. Graduated in IT from Mumbai University, skilled in user-centred design and building interfaces.
Ready to Work with Us?
Whether you are an armed force, a national-security agency, a government ministry, a critical-infrastructure operator or an integration partner, start with your operational problem and your security boundary. We will scope a bounded next step.