Defence Software Development Company for India’s Defence and Aerospace Programmes

Most defence software fails at accreditation, not in the demo: it assumed internet access the network does not have, a cloud the data may not leave, or a vendor licence the programme cannot carry. Discovery starts with the procurement and the boundary — the DAP 2020 category the programme sits in, its indigenous-content obligation, the classification of the data, the clearances the team will need — and settles the sequence: what to build, what to buy from an Indian OEM, what to leave alone. The output is a roadmap costed against accreditation risk, and a first release the unit or the plant will use.

Role-based operational applications for planning, tasking, reporting, inventory and administration, built to run on an air-gapped or restricted network with no dependency on a service outside it. Access follows the organisation’s roles and clearance levels; every action writes an audit record; deployments are reproducible from signed artefacts, because a patch may arrive on a disc rather than over a wire. We do not build weapon or fire-control systems; we build the enterprise and mission-support software around the programme, which is where most of the indigenisation work sits.

Readiness is a data problem before it is a spares problem. We build the models and dashboards that forecast component failure and spares demand from maintenance, inspection and sensor history, and turn them into schedules the depot can act on — with documented features, thresholds and review cadence, an explanation the maintenance officer can check, and drift monitoring in service. For a tractor OEM we built predictive maintenance across legacy and connected fleets — component issues detected 45% faster, unplanned-downtime risk down 35% — and the same pattern applies to vehicle, vessel and ground-equipment fleets.

Production, quality and traceability systems for licensed defence manufacturers and their MSME suppliers, built for the inspecting authority as much as for the line: lot and serial traceability, non-conformance and inspection records in the form DGQA or DGAQA samples, and indigenous-content accounting the contract can be audited against. Vendor portals bring suppliers listed on SRIJAN or won through Make-II into the programme with their documents and certifications. For a multi-location manufacturer we joined production, attendance and wages on one platform — wage calculation 55% faster — the same shop-floor discipline.

Inventory as a stream of events from receipt to issue, across depots, workshops and forward locations, with lot, serial and shelf life behind every count and reconciliation against the system of record. Demand signals feed the readiness models; supplier and indigenisation status sits on the same data; and a query about which lot went where answers in seconds rather than in a week of registers. Pipelines run inside the boundary with lineage recorded for every fact, so the auditor and the logistician read the same number.

Technical publications, maintenance manuals, standards, trial reports and decades of programme documents hold the knowledge, and finding the right page takes a day. We build source-backed retrieval and knowledge assistants that run entirely inside the network — no external model call, no data leaving the boundary — with access that respects classification and roles, and answers that carry their citation. For a global consumer-goods company we built the production-grade retrieval platform that made information 70% faster to find and cut manual knowledge discovery by 60%; the architecture is the same, the hosting is not.

Inspection of components, assemblies, welds and coatings at line speed; condition monitoring of equipment from imagery; and review workflows in which a person confirms what the model flags. We design the optics, the model and the reject signal, validate against the programme’s own defect library, and integrate the result with the quality system so a defect becomes an inspection record with the image attached. Models are retrained under change control and documented for the certifying authority.

Defence enterprises run long-service systems that were never meant to talk, on networks that must not touch. We build the interfaces that connect them safely: controlled, logged data exchange between classification levels, one-way transfer where the policy demands it, identity and access integrated with the organisation’s directory, and monitoring that reports a stopped feed before operations notices. Idempotent processing means a duplicate message never creates a duplicate record, and every exchange keeps its evidence.

Systems written a decade or more ago run programmes that cannot pause for a rewrite. An API layer goes in front of the legacy core first, so new applications, analytics and integrations ship beside it; modules are re-platformed one at a time — reporting, inventory, workflow — each run in parallel and reconciled before the old path is retired, with cutovers planned around exercise and inspection calendars and re-accreditation planned in rather than discovered. The organisation sees continuity; the team gets software it can change.

Segregated environments, signed artefacts, dependency and secrets scanning, software bills of materials, access reviews and incident records — produced by the pipeline as it runs, so the CERT-In, ISO/IEC 27001 and programme-security evidence exists because the work was done. Deployments are designed for disconnected delivery: a release that can be carried onto the network, verified and installed without an internet connection, and rolled back the same way.

Where the data may live is decided before the architecture is: on-premise and air-gapped for classified programmes, MeitY-empanelled Indian cloud for the rest, hybrid where a programme spans both. We design the environments, the identity model, the network segmentation and the resilience for each, with the residency and audit records the security accreditation asks for, and a cost model the programme office can defend.