Pradeep Kumar — Applied AI & Industrial Systems

The practice

I work with organisations whose operations have outgrown the systems that run them.

You know your operation better than any consultant will. What is usually missing is the engineering capacity to turn that knowledge into something that runs by itself. A typical engagement begins with a bottleneck — dispatch decisions made on intuition, reporting that takes three days, compliance evidence scattered across mailboxes — and ends with a system in production that people use every day. I stay responsible for that system after it goes live.

I work as a single senior practitioner. There are no junior layers billed to you and no handover chains: the person who designs the system is the person who builds and operates it. That limits how much I take on, and it is the reason the work holds together.

Background

In operational IT since 2002 — most of it running it rather than advising on it.

Technical support and server administration in telecommunications and banking, nine years at an international gaming technology company through to Head of IT, and four years as Chief Technology Officer of an Austrian group of companies.

Banking, logistics, iGaming, telecommunications, hospitality and IT services. That range is why the work transfers: a dispatch bottleneck and scattered compliance evidence are the same class of problem seen from two sides, and the pattern is only visible if you have stood on both.

English, German and Tamil.

If a bottleneck of your own comes to mind while reading this, write to me — a short description is enough to tell whether it is worth a conversation.

Capabilities

Six things I do, end to end.

Applied AI

Language-model assistants and agents, document and speech understanding, decision support, operational optimisation. Model choice driven by cost and reliability, not novelty.

Data & systems integration

Connecting ERP, production, logistics and financial systems into data foundations that can actually be queried, trusted and reported on.

Software engineering

Custom platforms, services and internal tools — from prototype through production hardening, authentication, role models and audit trails.

Infrastructure & operations

Cloud and on-premise architecture, virtualisation, monitoring, backup and disaster recovery, network design, incident response.

Security & compliance engineering

Hardening, penetration-test remediation, access and tenancy models, GDPR and EU AI Act implementation, audit-ready evidence and logging.

Advisory

Feasibility studies, architecture and code review, AI readiness assessment, build-versus-buy analysis, technical due diligence.

Selected work

Systems in production, not slide decks.

Transport & logistics — Austria

AI-assisted dispatch and route optimisation

Dispatchers assigned orders to drivers manually under time pressure. Assignment quality depended entirely on individual experience, and that knowledge was not transferable to new staff. I built a recommendation service reading live ERP and order data, combining real routing distances, driver capability profiles and customer-specific constraints — delivered as advisory output alongside the existing dispatch system, with a human always making the final call.

Result

Top-recommendation accuracy rose from 6% to 26% measured against the dispatchers' own choices, running on the client's own infrastructure at negligible marginal cost.

Customer operations — speech AI

Voice agent and call-intelligence pipeline

High inbound call volume for routine bookings, and years of recorded conversations holding operational knowledge nobody could reach. I built a voice agent handling structured requests through a purpose-built API gateway into the operational system, plus a transcription and speaker-diarisation pipeline that turned the historical archive into a searchable knowledge base.

Result

Over 3,000 historical calls processed into structured, searchable records; routine call handling automated with defined escalation to human staff.

Regulatory technology — highly regulated sector

Compliance automation across multiple jurisdictions

Operators in a heavily regulated industry must evidence compliance against divergent and frequently changing national rulebooks — work traditionally done by hand by expensive specialists. I built a multi-tenant platform with AI-assisted regulatory change monitoring, structured evidence management, tenant-level data isolation and audit-ready reporting. Human review is mandatory before any regulatory conclusion is issued.

Result

130 controls across 25 jurisdictions. Compliance assessment reduced from weeks of specialist work to a reviewable, evidence-linked report. Independently penetration-tested; all findings remediated.

Own ventures

Products I designed, built and run myself, outside client work. They are here because they show the same thing from a different angle: a complete system, taken from idea to production by one person.

Product · HR technology

PAIR

An AI-assisted HR platform for small and mid-sized employers. Product decisions, architecture, application, database, infrastructure and deployment pipeline — all of it mine, which is the point: it is the working proof of the end-to-end claim on this page.

Product · direct-to-consumer

பூ Made

A craft brand selling into a WhatsApp-first market in India. Catalogue, ordering and fulfilment built around how customers there actually buy, rather than around a standard webshop — a useful reminder that the right system depends on the market, not on the stack.

How I work

Five commitments that shape every engagement.

Working software early

Short, clearly scoped increments that reach production, rather than long specification phases that end in a document.

Accountable across the full chain

Feasibility, architecture, build, deployment, operation and maintenance — one point of responsibility, and it does not move.

Honest on build versus buy

Where a proven product already solves the problem, I say so instead of building. What you are paying for is judgement, not billable hours.

Cost as a design constraint

I work for organisations where budget is genuinely finite. Running cost is a design decision made at the architecture stage, not a surprise discovered in year two.

Your infrastructure, your data

Where data sovereignty matters, systems run on infrastructure you control. No lock-in to a platform only I can operate.

Responsible AI

Built to be explained, logged and overruled.

I treat governance as an engineering requirement, not a policy annex. Systems I deliver are designed to be explainable to the people who depend on them, to record the decisions they make, and to keep a human in the loop for consequential outcomes.

GDPR obligations and EU AI Act requirements are implemented as part of the build rather than documented afterwards. Where sovereignty matters, data stays on infrastructure the client controls.

Contact

Let's talk about the bottleneck.

Emailpradeep@pradeepkumar.eu

Telephone+43 677 61696136

LinkedInlinkedin.com/in/pradeep-kumar81

AddressHausfeldstraße 65b/2
2232 Deutsch-Wagram, Austria

Capability statement PDF · 2 pages · 88 KB