Seven platforms. One stack.
We build the origination, evidence, decision and compliance layer for Indian financial services. Our products read a business or a borrower from evidence others do not collect, decide with a named human in the loop, and keep the labelled outcome as the asset that compounds.
The stack
Read front to back, it is one system.
A front door finds the customer. Three evidence engines answer three questions: what is on paper, what is the real income, and can the intent be trusted. An operating system turns the three reads into one decision. Compliance and simulation sit underneath.
The principles
The rules every product is built to.
The products differ. The principles do not, because a bank, a regulator and an investor will test us on them.
Intelligence recommends, a named human decides
Nothing of consequence is filed or disbursed autonomously. That is what the regulator requires, and it is the product.
Own the data
Every decision, good or bad, is a labelled example kept for the next one. The labelled outcome data is the moat.
Create evidence, do not re-process records
The edge is in reading what others cannot read at the point of origination, not in re-processing records everyone already holds.
Build it inspectable
A buyer’s security team can read the product. Several platforms are single files by design, so the data does not leave the browser.
Claim only what is there
A deterministic engine is called a deterministic engine. A behavioural signal is called a behavioural signal. Defensible language is a competitive asset.
Asset light by design
We build the software and hold the intelligence and the data. The balance sheet risk sits with the lender.
Origin8 Revenue OS
Autonomous revenue infrastructure that runs the whole cycle, from finding a prospect to the login and the funded product.
Live at origin8.litehouseiq.com
What it is
An artificial intelligence agent orchestra for origination. Where a conventional aggregator ends at the handover of a lead, Origin8 carries the customer through qualification, engagement, login and know your customer, product match and the underwriting hand-off, and then learns from the outcome.
How it works
- Signals and prospect. Intent and first party data detect a prospect worth pursuing, and ideal customer cohorts are built and targeted.
- Engage and convert. Outreach runs across four channels, and the funnel is optimised continuously against outcomes.
- Login and match. The one time password and know your customer steps are orchestrated, and the lender or product is fitted to the applicant.
- Fund and learn. Underwriting is handed to Income360, and every outcome retrains the models, closing the loop.
Where it fits
The top of the funnel. It feeds the evidence engines and the operating system, and everything it learns about a cohort feeds the next targeting cycle.
DocLens
Document intelligence for Indian lending that runs inside the lender’s own network.
Live at doclens.litehouseiq.com
What it is
DocLens reads the documents in a loan file, extracts the fields an underwriter uses, derives the credit analytics, and flags tampering. Recognition engines are commodities. The India specific layer above them, the statement layouts, the defensible derivations and the forgery forensics, is the product, and it deepens with every file.
How it works
- Ingest and classify. Each file is taken in and classified across seventeen document types.
- Extract and validate. Fields are extracted and checksums validated on capture, from the Aadhaar check digit to the GSTIN.
- Analyse. Bank statements are reconciled row by row, salary is detected by regularity, and obligations and average balances are derived.
- Check and route. Ten forgery checks run, and the file is routed straight through, to review, or to refer.
Where it fits
Every extracted value carries its page, its source text, its method and a confidence. That provenance is what lets a bank accept an automated read of a file.
Multimodal DeepTech
Assembles an interactive Customer 360 from every signal a borrower brings, and returns a trust and affordability read an underwriter can act on.
Live demonstration, access on request
What it is
A multimodal intelligence layer that assembles a single, interactive Customer 360 from documents, transactions, the public rails and the field record. It reads the whole customer rather than one data point, and returns a trust and ability signal created at the point of origination.
How it works
- Assemble. Every signal the customer brings is drawn into one interactive Customer 360 profile.
- Corroborate. Each attribute is cross checked against the consented data rails, so a figure resolves to a source rather than a claim.
- Score and explain. The signals fuse into one trust and affordability read with an explainable trail, for a named human to act on.
Where it fits
Paired with the ability read from Business X-Ray inside Income360, it completes the picture an underwriter needs.
Business X-Ray
Rebuilds turnover, margin and the working capital cycle for a business that keeps no accounts.
Live at bxray.litehouseco.com
What it is
The credit is not bad. The evidence of it is physical, and nobody has built the means to read it. Business X-Ray reads the shop, the stock, the machinery, the people at work and the behaviour of the owner, and converges them with GST and banking data where they exist into one assessed income, with a confidence band attached.
How it works
- Guided capture. The trade is chosen from fifty six profiles, and photographs are captured live from inside the field agent geofence, each read on the spot and confirmed by a person.
- Scenario reconstruction. A conservative, a base and an optimistic turnover are produced, and the interval narrows as evidence arrives.
- Cash flow rebuild. A monthly cash flow is rebuilt bottom up, and every line carries a score for how well it is corroborated.
Where it fits
The ability read inside Income360. It is triangulated with the Customer 360 read, the rails and the field before a named credit officer decides.
Income360
The origination to servicing operating system, from the field agent’s first contact on the phone to the last instalment.
Live at psl.litehouseco.com
What it is
Where the evidence engines, the scorecards, the credit engine, the field roles and the compliance spine come together into one loan lifecycle. If the evidence engines answer the parts, Income360 is the whole.
How it works
- Credit engine. A bureau and affordability model turns the reads into one decision to approve or decline.
- Hard gates. Clear title and a minimum distress value are enforced regardless of score, alongside obligation and loan to value caps.
- Credit memo. An eight section appraisal memo fills as the case advances, and a named human signs it.
Where it fits
The assembly point of the whole stack. Every score resolves to a source, every decision to a named human, and every file to a labelled outcome that improves the next.
Ikigai
The compliance, credit and audit operating system for India’s regulated financial institutions.
Live at litehouseiq.com
What it is
A continuous compliance system: a deterministic policy engine, continuously recomputed, that keeps the institution current with its filings and its prudential floors across RBI, SEBI, NHB and DPDP, and across GST, income tax, PF, ESI, PT and company law. It is not an agentic system, and we do not describe it as one.
How it works
- Regulatory library. Every instrument from fifteen issuing authorities, each with a reference card.
- Submission universe and data room. Returns scoped to the entity type, fed from one canonical repository in the automated data flow shape.
- Prudential rule engine. Policy as code. A breach blocks the compliance officer’s sign-off until it is justified by name.
- Regulator format output. XBRL, tax network and statutory files generated for the officer to upload. The Praman layer runs the DPDP lifecycle over WhatsApp.
Where it fits
The compliance layer under the whole business. Nothing files autonomously: a named human signs every submission, a different human checks it, and the chain is hashed.
TwinSim
A synthetic population digital twin of Indian financial behaviour. Test it on the country before you test it on a customer.
Live at twinsim.litehouseiq.com
What it is
Statistically faithful synthetic citizens, built from public and licensed survey distributions, so a product, a price, a message or a policy can be tested against how the country actually behaves before it reaches a real person.
How it works
- The synthetic population. Resolved from a single household up to the nation.
- The orchestration. A conductor routes each task to specialised agents for compliance, credit, fraud, reporting, customer intelligence, insurance, microfinance and market simulation.
- The arbitration. Compliance and fraud hold veto authority, and ties or low confidence escalate to a human.
- The ledger. Every action, model version and human override is written to an append-only record.
Where it fits
It lets a portfolio built on the rest of the stack be stress tested before it is written, and its governance native design is the reference for how a regulated institution runs agents at all.
Jugaad
A privacy-first, on-device AI assistant built for how Indian households and small businesses actually work.
Live at jugaad.litehouseiq.com
What it is
An instinct-type assistant that meets a user in their own language and their own way of working, rather than forcing them onto a form. It keeps the data on the device, so what is private stays private.
Where it fits
The front-line assistant across the stack, from a borrower asking a question to a field officer needing an answer, with the same privacy-first design underneath.
MyVault
A sovereign, local-first digital-twin operating system for any professional and any organisation.
Live at myvault.litehouseiq.com
What it is
Two separate vaults, work and personal, each with its own keys, that build a structured digital twin of the user and run role agents to turn scattered information into decision-ready, auditable action, on mobile and desktop, without the data leaving the user’s control.
Where it fits
The sovereign memory layer under the stack, so a professional or an institution keeps a second brain that acts, shaped to the role, with nothing leaking to someone else’s cloud.
End to end
One loan, across the whole stack.
The platforms are meant to be seen not as seven products but as one loan moving through them.
Find and log in
Origin8 detects a prospect, engages across four channels, and carries the login and know your customer step.
Read the documents
DocLens reads the statements, the KYC, the GST and the tax returns, with a provenance trail and a tampering check.
Read the business
A field agent captures the premises, and Business X-Ray rebuilds turnover, margin and cash cycle with a confidence band.
Build the Customer 360
Multimodal DeepTech assembles an interactive Customer 360 from every signal and returns a trust and ability read.
Decide
Income360 assembles the three reads, runs the scorecards and the hard gates, confirms every response against the data backbone, and a named credit officer signs.
Stay compliant
Ikigai keeps the entity that lends current with its filings and its prudential floors.
Test before you write
TwinSim stress tests the portfolio, the price and the policy before the next cohort is written.
Why it compounds
Three things a competitor cannot buy.
A competitor can copy a screen. It cannot copy the labelled book underneath it, the channel, or the trust.
Labelled outcome data
Every loan adds a labelled example across images, documents, transactions and repayment. It exists only because the loans are written.
The distribution channel
Cooperative societies and their agents already meet these businesses every day, a reach that cannot be assembled from outside.
Governance native design
Human in the loop, veto for compliance and fraud, and an append-only audit trail. This is what lets a bank and a regulator trust the decision.
Talk to us
Ask for a walkthrough, or a free month on your own files.
Tell us which platform and what you would want to see. The form opens an email to our customer care desk, so nothing is stored on this website.