Nokshi Technology
An applied artificial intelligence studio, engineering production systems where precision is not optional.
Nokshi Technology designs and delivers systems across three disciplines — artificial intelligence product engineering, applied machine learning, and AI strategy and governance — for regulated industries, public-affairs and advisory firms, clinical and research institutions, and mid-market enterprise where the cost of a misjudged system substantially exceeds the cost of building the right one.
Founded by Dr Rittick Barua
— PhD, University of Cambridge
·
MEng (First Class Honours), University College London
·
co-author, UK Government DSIT whitepaper on cyber-security risks to artificial intelligence (2024)
·
research-engineering affiliation, UCL — Royal Free Hospital (cell-spheroid segmentation, validation intersection-over-union 0.968).
Representative engagement
Case studies describe live engagements. We publish one when the underlying work is substantive enough to warrant it, and when the client is content for their name to be attached.
Rud Pedersen Group
Chat:R
A white-label AI platform for a political consultancy — private, tenant-bound, and serving the full staff on an extended LibreChat fork.
-
~1,000
staff served across 3 environments
-
$210/mo
infrastructure, down from ~$400 on the original AKS design
-
3 custom features
Projects, Agent linking, SharePoint MCP gateway — built on top of LibreChat
Disciplines
Three disciplines, engaged separately or in combination.
Most engagements are fixed-scope deliveries inside one of the three disciplines below. A smaller number of engagements span all three, typically where the underlying problem involves model development, product delivery, and governance in the same project.
- 01
AI product engineering
Full-stack delivery of language-model platforms, retrieval-augmented knowledge systems, agent frameworks, and bespoke AI interfaces — deployed inside the client's tenant, with data architecture designed to keep it there. Reference build: Chat:R, an extended LibreChat platform for a European political consultancy.
Learn more - 02
Applied machine learning
Computer vision, time-series forecasting, and interpretable predictive modelling for clinical, operational, and financial settings. Methods are chosen by problem rather than by trend. Reference build: a U-Net segmentation model for high-throughput cell-spheroid assays at UCL Royal Free Hospital (IoU 0.968, Dice 0.983).
Learn more - 03
AI strategy and governance
Published advisory, programme design, and readiness assessment — informed by peer-reviewed and government-published work, including a UK Government DSIT whitepaper on cyber-security risks to artificial intelligence (Barua et al., 2024) and a Jersey Finance–published guide to artificial intelligence in Jersey's finance industry (McCay and Barua, 2024).
Learn more
A fuller description of each discipline, with representative deliverables and engagement shapes, is on the Services page →
Method
Every engagement follows the same three-phase shape: diagnose, build, embed.
The shape is deliberately simple. The purpose of keeping it consistent is to make the scope, the deliverables, and the handover criteria legible to the client from the outset — and to produce a system the client's own team can operate once the engagement concludes.
- 01
Diagnose
A short, paid discovery phase. We map the underlying constraints — the data, the tenancy, the team, and the operating context — and return a written architecture and delivery plan that the client's own engineers can review and challenge.
- 02
Build
A fixed-scope build, typically six to fourteen weeks. Infrastructure, continuous integration, monitoring, and security review are treated as first-class scope rather than as deliverables appended at the end. The intention is a system that is usable on the day it is handed over, not a demonstration that is subsequently rebuilt.
- 03
Embed
Handover is a design property of the engagement, not a concluding step. Documentation, environments, and architectural reasoning are written down so that the client's team can continue to operate and extend the system. An iteration retainer is available where useful, but permanent dependency is not the objective.
Typical clients
Four environments in which precision, provenance, and compliance tend to dominate the technical brief.
In each of these settings, the underlying constraints shape the architecture rather than the other way around. Applied artificial intelligence becomes useful in these contexts only when the system is designed around those constraints from the beginning.
-
Regulated industries
Financial services, healthcare, and legal organisations in which data residency, tenancy, and audit posture form part of the technical brief rather than appearing at the end of it.
-
Public affairs and advisory
Political consultancies, policy shops, and advisory firms handling client briefings and positions that cannot be placed into general-purpose consumer AI tools.
-
Clinical and research institutions
Hospitals, universities, and research groups applying machine learning to segmentation, imaging, and evaluation problems — with the ethical and publication requirements those settings attach.
-
Mid-market enterprise
Established organisations moving past proof-of-concept artificial intelligence and into systems that must integrate with existing data estates, identity providers, and governance processes.
Named-author publications
Publications carrying the studio's positioning.
The two 2024 publications below address applied artificial intelligence directly and are the institutional anchoring most readers of the site will find useful as a starting point.
UK Government · DSIT · 2024
Cyber security risks to artificial intelligence
McCay, Al-Khalidi, Peng, Crossman-Smith, Barua
Read on gov.uk (PDF)Jersey Finance · 2024
Guide to Artificial Intelligence in Jersey's Finance Industry
McCay & Barua
A fuller list, including earlier peer-reviewed work, is on the Publications page →
From the founder
"Before a written proposal tends to be useful, there is usually a short conversation that clarifies what the actual problem is — and, on occasion, whether it is a problem I am the right person to help with. I generally prefer to begin there."
— Dr Rittick Barua, Founder