Applied AIResearch, models and products under one roof

Every industry is about to run on intelligence. We build that layer.

We research it, we build it, and we help businesses and consumers actually use it, from datasets and models through to finished products, across healthcare, finance, agriculture, law, education, logistics, manufacturing, retail and government.

Any industryhealthcare and finance through to agriculture and government
However you buycloud, your servers, or inside your own network
0 productsalready in the field, with more behind them
End to enddatasets and models through to finished product
Shipping now

Eight products,
and counting.

These are not the plan. They are what the plan has produced so far. Some are bought by an institution, some are opened by one person on a phone. All of them sit on the same stack, with the same audit trail running underneath.

Early access

Atlas

Everyday money

Reads a loan or a policy and returns what it costs across the full term.
Early access

Aureus

Tax and compliance

Runs a firm's compliance calendar end to end, and tracks the rules as they change.
Early access

CounselOS

Legal practice

Runs the matters, dates and papers of a practice, from first meeting to final order.
In pilot

Medperita

Hospitals

Builds and marks clinical competency assessments, with a different paper per candidate.
In development

EduNexus AI

Schools

Admissions, attendance, fees, timetables and results, running as one school system.
In development

EduNexus Learn

Students

A tutor on the student's phone, working from their own board and syllabus.
In development

TerraIndex

Property

Assembles the full picture of a locality before you commit to an address.
Nearly ready

Jyotish Engine

Heritage

Classical Jyotish computation, in ten Indian languages.

What we build

Every layer,
built in one place.

We work the full height of the stack, because the interesting problems live in the joins: between the data and the model, between the model and the process, and between the software and the person who has to sign for it.

01Data and research

Ingestion pipelines, proprietary datasets, benchmarks, evaluation frameworks and synthetic data generation. The unglamorous half of AI, and the half that decides whether anything above it is trustworthy.

02Models and IP

Training and fine-tuning, model weights, algorithms and the patents and trade secrets around them. What we learn on one problem becomes an asset we own and can carry into the next.

03Infrastructure

Deployment and inference, agents, orchestration, integration and middleware: the machinery that gets a model out of a notebook and into a process that runs every day without anyone watching it.

04Platform

One codebase, delivered as software, as a platform, as intelligence over an API, or installed on hardware you own. How you buy it should not change what you get.

05Products

Co-pilots, decision support and finished software built for a named job in a named industry, where the whole stack below finally becomes something a person can open on a Monday morning and use.

Capabilities

The disciplines
we work in.

A set of disciplines we assemble differently for every engagement, and add to whenever a problem demands something we do not yet do. Some jobs need one of these. The ones worth taking on usually need five at once.

Workflow automation

Robotic process automation, workflow orchestration, business process management, task and pipeline automation. Standalone, or with a model in the loop where the rules run out.

Agents and LLM systems

Retrieval, tool use, multi-step agents and large language model integrations that hold up against a real corpus, a real schema and a real deadline.

Computer vision

Document understanding, inspection, detection and measurement, reading the things a process depends on that were never typed into a system in the first place.

Natural language

Extraction, classification, summarisation and question answering across the languages and registers actually used in Indian records, contracts, prescriptions and filings.

Predictive analytics

Forecasting, scoring, assessment and recommendation, delivered as a number a team can act on, with the reasoning attached rather than buried.

Generative and multimodal

Systems that draft, compose and produce across text, image, audio and structured output, held to the same evaluation bar as everything else we ship.

Datasets and evaluation

Proprietary corpora, benchmarks and evaluation harnesses, built because the public ones do not cover the domains we work in, and licensable in their own right.

Integration and middleware

The connectors, adapters and translation layers between the systems a company already owns, and usually the reason an automation project succeeds or quietly dies.

No-code and low-code

Surfaces that let the people who understand the process change it themselves, without a ticket, a release cycle or an engineer in the room.

Industries

Wherever the work is,
one way of working.

These are the industries we are in today, and the list is not closed. The method travels further than the domain knowledge does. Learn the process from the people who run it, build against the rule as written, and leave the last decision with a person.

Delivery

However you need
to buy it.

As many ways to take delivery as you have constraints. Hospital records and legal files frequently cannot leave the premises, and that is a normal way to buy from us, not an exception we grudgingly support.

01 · SaaS

Software as a service

We run it, patch it and keep it current, with storage pinned to a region you choose.

02 · PaaS

Platform as a service

Build your own workflows and surfaces on top of our orchestration, models and connectors.

03 · AIaaS

AI as a service

Inference, extraction, scoring and generation as managed capability, without owning the infrastructure.

04 · API

API access

The same capability behind a documented interface, metered and versioned, for your own product.

05 · On-premise

Your own hardware

The identical build installed inside your walls, including networks with no route out. No reduced edition.

06 · Hybrid

Hybrid

Sensitive data and inference stay inside; everything that does not need to be there runs managed.

Subscription Perpetual licence Usage-based Outcome-based Dataset and model licensing Something else? Ask us
How we work

Four commitments
that do not move.

01

The last step stays with a person.

Automation prepares, drafts and queues. Sending, filing and signing stay with your team, and nothing crosses that line on its own. We are measured on how much we remove before that line, never on how much we do past it.

02

The work goes to the data.

Where the software runs is your decision, not a pricing tier. The same build installs on your hardware, inside your network, with no capability held back for the cloud edition.

03

Every answer carries its source.

Each figure, date and clause we return points back to the document, section or statute it came from. If a reviewer cannot get from the output to the origin in one click, we treat it as unfinished.

04

It says when it does not know.

An abstention a reviewer can act on is worth more than a confident answer they have to check. Where the source is silent or in conflict, the system flags it and routes it to a person.

Contact

Write to us, and a
person will read it.

Tell us the process you want finished and where your data has to stay. We will tell you honestly whether something we already run covers it, or whether it is a build.

Thank you. We have your message and a person will reply.

No newsletter, no sequence. One reply from someone who works here.

AISS Research

Intelligence you can
sign your name to.

Datasets, models, platform and products, all carrying the same mark, built under one roof and pointed at real work.

Start a conversation See the eight products