Everything we get asked, in one place

KriyaXlabs is an AI product studio based in Mangalore, Karnataka, India, founded in 2025. It operates two ways at once: it runs a portfolio of seven of its own AI products across education, design, assessment and architecture, and it takes on client engagements building custom web, mobile, AI and hardware products. The same engineering team does both.

KriyaXlabs, one word, capital K and capital X. It is also written KriyaX Labs, KriyaXLabs and Kriyax Labs in various places. They all refer to the same company at kriyaxlabs.com.

The office is at Door No: 13, Vertex Workspace, Gateway Building, MG Road, Ballalbagh, Mangalore 575003, Karnataka, India. We work with clients remotely from there, anywhere in the world.

KriyaXlabs was founded by Amogh in 2025. It is a GST-registered business in India, run as a proprietorship rather than a private limited company or LLP.

Both, deliberately. The product portfolio is where architecture gets proven under real load, and client work is where that architecture gets applied. Studios that only do client work rebuild the same patterns for every customer; product companies without client work see one problem shape forever.

Seven: Crampad Enterprise, Crampad, SketchXFlow, ExamXFlow, EduXFlow, ArchXFlow and PAAC. They range from shipping to unannounced. Client projects, such as WorkPay, are listed separately.

Yes. There is a screen-free device in development alongside Crampad, built once for both homes and classrooms and designed to be best in class at what it does. We have not described it further and will announce it on LinkedIn first.

Not yet. PAAC is unannounced. The first unit is called PAAC-1. It will be announced on LinkedIn before anywhere else.

Book a free thirty-minute discovery call at https://cal.com/kriyaxlabs, or email hello@kriyaxlabs.com. We reply within one business day. The office keeps Mon–Fri, 10:00–19:00 IST.

Yes. Work is remote from Mangalore, and we work with clients anywhere. Calls are scheduled around your timezone rather than ours.

Within one business day for email and enquiries. After a discovery call, a detailed written proposal follows within two business days.

Yes, and it is a common request. We start with a paid architecture review covering data model, tenancy, access control, build health and performance. You get a written findings document with fixes ordered by risk, and you keep it whether or not you continue with us.

Yes — it is the thing we do most. Every enterprise table carries an organisation id with row-level isolation, roles are scoped to the organisation and carried in the token, and org units can nest for multi-campus or multi-department customers. We run this architecture ourselves in Crampad Enterprise.

Server rendering for search visibility, static generation where content is stable, and a single TypeScript codebase from database to browser. It is also what every one of our own web products runs on, so we are not learning it on your budget.

Yes. We run a short audit first — usually two to three days — covering architecture, technical debt and risk, then propose either incremental improvement or a rebuild with the reasoning attached.

An MVP with real tenancy and roles typically ships in six to eight weeks. A marketing site with full structured data, two to three. A platform with multiple persona consoles is scoped per engagement after discovery.

It depends on who maintains it after us. Flutter if your team is new to mobile or values a single rendering engine across platforms; React Native if you already have a React web team and want to share skills and some code. We ship both and will tell you which fits, not which we prefer.

Yes, including the parts that usually cause rejections — privacy manifests, permission strings, data-safety forms and billing compliance. We hand over a release process your team can run afterwards.

Yes, and for most Indian consumer and education use cases it has to. We design for local persistence and conflict-aware sync from the start rather than bolting it on.

Gemini on Vertex AI, Claude, and OpenAI models, chosen per task. We routinely generate with a fast, cheap model and verify with a stronger one, because verification output is small and the asymmetry makes it nearly free.

A hard spend meter in currency that halts a run the instant it is crossed, a kill switch that refuses every billable call without a deploy, and measured per-run cost reported rather than estimated. These ship with every system we build.

Yes. We ship one in production for SketchXFlow, so coding agents can design, inspect and iterate programmatically. The same approach exposes your internal tools to agents safely.

Yes — retrieval over your documents with vector search, with attention to chunking, ranking and citation so answers can be traced to sources.

We design around the electronics — board placement, sensor sightlines, cable routing, thermal and assembly — and specify the parts. Custom PCB design is scoped per project with a partner where needed.

A parametric CAD model, STEP output for manufacture, print-ready files for prototypes, assembly documentation, a parts list, and photoreal renders.

Through to a manufacturable design and a prototype run, yes. Production tooling and volume manufacturing are handled with a manufacturing partner, which we can help you select.

Yes. The pipeline infers the layout structure the designer implied — flex, grid, breakpoints, components — rather than emitting absolute positions, and verifies the result by building it before you see it.

It is written to be read in a pull request, which is the constraint that makes it maintainable. It lands in your own repository, and your team edits it like any other code.

It removes the least interesting work — translating a settled design into structure, styles and breakpoints. It does not replace product judgement, state management or backend work, and we do not sell it as if it did.

We price by scope rather than hourly. Typical MVP sprints start in the low lakhs INR — mid four figures USD — for a four to eight week engagement. Every quote is bespoke, shared within two business days of a discovery call.

Most MVPs ship in four to eight weeks. The variable is scope, and we will tell you in the first call what has to be cut to hit a date.

A large share of our MVP work is with first-time and non-technical founders. We translate the idea into a concrete spec, architecture and sprint plan, and keep the whole engagement in plain language.

You do. Full ownership transfers on final payment; we retain rights only to pre-existing reusable components and general know-how.

Usually two to three days, followed by a written findings document with severity and effort against each item and a walkthrough with your team.

Yes, for a defined period, working in your repository and your process. This is billed as monthly retained capacity.

Yes, before receiving any confidential material. Send yours or request ours.

Fixed-scope projects, monthly retained capacity, and long-term product partnerships with custom — sometimes hybrid or equity-linked — terms for the right fit.

Very few at a time. We run our own portfolio with the same team, and the partnership model only works if we can give a partner that level of attention.

Exit terms are agreed at the start and leave you with everything — code, infrastructure, documentation — so leaving is a decision, not a hostage negotiation.

Schools, colleges and universities. The institution is the billing and policy boundary, with optional campus or department units beneath it, so multi-branch colleges and districts model cleanly.

Yes. OIDC single sign-on with domain routing, and SCIM 2.0 provisioning for users and groups, are built into the platform and disabled by default. SAML is available on request.

Every enterprise table carries an organisation id with row-level isolation, and the inherited content tables were retrofitted with the same column. Roles are org-scoped and carried in the JWT for the active organisation.

Something with no screen. That is the extent of what KriyaXlabs has said about it. It is in development alongside the Crampad app, is designed to be best in class at what it does, and will be announced on LinkedIn first.

Yes. It is being built once for both. The same device serves an individual learner at home and a classroom under an institution's Crampad Enterprise account.

Indian exam aspirants — primarily NEET and JEE candidates in Class 11, Class 12 and droppers, with Class 9 to 12 NCERT board students as a secondary audience.

Yes. Beyond generating the design, SketchXFlow converts it into production code for Next.js, Flutter or React Native, builds it in an ephemeral cloud container to verify that it compiles, repairs failures automatically, and pushes the finished project to a GitHub or GitLab repository.

Multi-page websites, application screens, advertising creatives for Meta, Instagram and Google display placements, posters, flyers, social posts, and complete logo and brand identity kits.

JEE Main, NEET and state CET, with paper blueprints for JEE Main, JEE Advanced and NEET-UG. A computer-based testing branch is in build for JEE Main and NEET.

A second, stronger model independently solves every generated question rather than merely reviewing it. No question ships unless exactly one option is provably correct.

No. EduXFlow is deliberately not an LMS — there are no courses, no enrolment and no completion tracking. It is an open video platform with a feed and discovery, scoped to educational content.

It is education only, and it ranks by learning value and completion rather than by watch time. Subject and level are first-class metadata, so browsing by topic is real navigation rather than a tag search.

Site plans, floor plans, elevations, sections, working drawings, construction details, structural drawings including foundation and framing layouts, concept sketches, full written specifications, renders, and a parametric 3D model exportable as glTF.

From live sources rather than assumptions: geocoding and elevation for terrain, a ten-year daily climate archive for design temperatures and wind, roughly three decades of flood reanalysis at the exact coordinates, the USGS earthquake catalogue for seismic exposure, and OpenStreetMap for neighbouring building heights.

PAAC is an unannounced product in development at KriyaXlabs. Details have not been released. Announcements will be made on the KriyaXlabs LinkedIn page.

Next.js 16 with the App Router and React 19, on TypeScript. We upgrade our own products first, so we hit the breaking changes before a client does. Existing projects on the Pages Router can be migrated route by route rather than in one rewrite.

Yes. We start with a two-week architecture review covering routing, data access, tenancy boundaries, build times and Core Web Vitals. You get a written findings document and a prioritised plan whether or not you continue with us.

Usually, though not always. We run Next.js on Vercel, Cloudflare and Google Cloud depending on where the data has to live and what the compliance picture looks like. The framework does not force the hosting decision.

Usually with Next.js, because routing, server rendering and image handling are solved problems there. We build standalone React applications when the front end must be embedded in an existing host application or shipped as a widget.

Server state stays on the server wherever the framework allows it. What remains is genuine client state, which is small enough for React context and reducers in most applications. We add a state library when the application proves it needs one, not by default.

Flutter when the interface is custom, animation-heavy or must look identical on both platforms, and when offline behaviour matters. React Native when you already have a React web team and want to share people and patterns between web and mobile. We build both and will tell you which one your situation argues for.

Yes, under your developer accounts. We prepare the listings, privacy declarations and data-safety forms, run the staged rollout, and hand over the accounts and signing keys at the end of the engagement.

Four to eight weeks for most scopes, run as fixed-scope sprints with weekly builds you can install. Offline-first requirements and payment integrations are the two things that most often push it toward the upper end.

Yes. We take embedded engagements where our engineers join your standups, work in your repository and follow your review process. That is the most common shape for React Native work, because the codebase usually already exists.

Yes, under your accounts, including the privacy and data-safety declarations. Payments apps attract extra scrutiny from both stores, so we prepare those answers before the first submission rather than after a rejection.

REST by default, because it is easier to cache, log and reason about across many client surfaces. We build GraphQL where a client genuinely needs to shape its own queries, most often when several very different front ends read the same domain.

Usually, and it is one of the most common reasons companies come to us. It is a data-model migration before it is a code change, so we start with a schema and access review and give you a written plan with the risky parts named.

Mostly integration, orchestration and verification around hosted models, which is what the great majority of products actually need. We are direct about the line: if your problem genuinely requires training a model from scratch, we will say so rather than sell around it.

Batching, caching, model tiering and hard budget ceilings enforced in code. We instrument cost per operation from the first week, because a pipeline that is affordable in a demo and ruinous at volume is the most common way AI products fail.

For many, yes, particularly when row-level security is used properly. Where a customer requires data residency, private networking or their own single sign-on, we run PostgreSQL on the cloud they mandate and keep the same schema and access model.

Yes. A database review covers the schema, row-level security policies, index coverage, slow queries and migration hygiene, and comes back as a written document with fixes ordered by risk.

No. We build on Vertex AI and Gemini most often, and also on Claude and OpenAI models. The integration layer is written so the model behind a task can be swapped without rewriting the product around it.

A separate verification stage that evaluates output rather than trusting it, plus retrieval that grounds answers in your own content, plus explicit refusal paths. We also keep and count the failures, because a system that never reports a rejection is not checking anything.

Yes. We deploy into your project with your billing, your quotas and your service accounts, and hand over infrastructure-as-code at the end so nothing depends on us remaining involved.

We build learning platforms, and we start by asking whether you need one. Where an existing LMS covers the ground, the useful work is usually integration and the surfaces around it. Where the institution's model does not fit an off-the-shelf product — unusual org structures, assessment rules, or licensing — a purpose-built platform is justified, and that is what Crampad Enterprise is.

Yes. OIDC single sign-on and SCIM 2.0 user and group provisioning are already built in Crampad Enterprise and switched on per institution. Most school and university procurement processes ask for both.

Data scoping is enforced at row level, audit logs record every mutation including our own staff actions, and data handling is documented per deployment. We will work to the specific regime your institution answers to rather than a generic policy.

With a separate pass that judges the generated item rather than the model that produced it, checking the stem, the key, the distractors and the alignment to source. Items that do not clear the bar are rejected and recorded. We wrote up the approach in our article on verifying AI-generated exam questions.

Yes. ExamXFlow already has a CBT branch, so the delivery mechanics — timing, sectioning, resumable sessions, result reporting — exist rather than needing invention. What varies per institution is the item bank, the paper structure and the reporting.

We integrate proctoring rather than building surveillance from scratch, and we will be straightforward about the trade-offs. Most integrity problems are better addressed with item randomisation, banking discipline and delivery design than with cameras.

We work with structured output formats and integrate with the tools a practice already runs rather than asking it to move. What is feasible depends on which tools and which parts of the workflow, so the first step is a scoping conversation about the actual pipeline.

It is in build. If you are an architecture practice interested in the problem, a conversation now is useful in both directions — talk to us and we will tell you honestly where it stands.

No. We build software; the regulated functions sit with licensed providers. We integrate with your gateway, processor or banking partner and build the product around it.

We integrate against whichever provider your business and jurisdiction require, and we design the integration layer so a provider can be changed without rebuilding the product around it. That abstraction matters more than any particular vendor.