Production architecture for insurance, energy, and trading systems.Production architecturefor insurance, energy, and trading systems.

Elysian Systems · Principal architecture and production engineering

We design and operate regulated, real-time, and data-heavy systems where correctness, risk, and uptime matter.

Insurance modernization, Solvency II / IFRS 17 frameworks, Python data platforms, AI-assisted engineering, and live automated trading infrastructure.

Dušan Krovinović

Former CIO · enterprise architect · principal engineer · 30 years

Available · B2B

Discuss an engagement View systems

B2B contracts
Remote EU/worldwide
Insurance modernization
Python data platforms
Principal architect
Fractional CTO
Former CIO · 30 years

Engagement model

For organizations modernizing systems where domain rules, legacy platforms, data, risk, and production constraints all collide.

Core offer

Designed for

Policy administration, underwriting, billing, claims, commissions, reinsurance, Solvency II / IFRS 17, actuarial and insurance data platforms

Insurance modernization architecture

Elysian Systems supports modernization of life insurance platforms, regulatory frameworks, actuarial-adjacent systems, claims, commissions, reinsurance, and insurance data platforms while protecting operational continuity.

Discuss insurance modernization

Designed for

Energy and trading platforms, market and time-series data, forecasting, valuation, P&L, risk controls, PostgreSQL, Azure, Kubernetes

Energy, trading, and data platform architecture

Elysian Systems designs Python-led platforms for energy and trading domains, including market data, forecasting, valuation, P&L, production ML, live execution infrastructure, and cloud-native operations.

Discuss a data platform

Designed for

Architecture governance, modernization roadmaps, vendor decisions, team guidance, AI-assisted engineering adoption

Fractional CTO / principal architecture

Elysian Systems provides senior architecture governance, technical strategy, vendor and platform decisions, modernization roadmaps, AI-assisted engineering adoption, and mentoring for product and platform teams.

Discuss advisory

Systems in operation6 systems

Core insurance and regulatory platforms

Life policy administration · Solvency II · IFRS 17

production

Long-term ownership and evolution of a core life insurance administration system on Amarta/FIS, supporting 100K+ life policies across traditional, unit-linked, annuity, endowment, and risk products. The work covered policy servicing, premiums, claims, commissions, reinsurance, underwriting, data warehouse workloads, and related operational processes. Separate C# framework work provided the technical foundation for Solvency II / IFRS 17 applications, with actuarial calculations implemented by actuarial teams.

Context: Live life administration system · regulatory framework deployments · 100K+ policy administration scale
Role: Principal architect · core system ownership · C# regulatory framework design

Core modules

Full core scope

100K policies

100K+ policy administration

Energy trading and risk data platform

Python · energy · gas · futures · P&L · risk

production

Python-based data platform for energy portfolio, gas, futures, and supply operations, designed and implemented end to end by Elysian Systems. The system supports imports, prices, trades, hedging, mark-to-market, realization, company/book/trader calculations, P&L reporting, risk limits, operational controls, and notifications through a repeatable production calculation process.

Context: Live · energy trading operations · supply P&L · risk and limit management
Role: Architecture · Python implementation · data warehouse and calculation engine ownership

Python-built

Data warehouse and calculation engine

P&L + risk

Hedging, limits, reporting

Flexibility and industrial energy platform

Azure · Event Hub · IoT · Kubernetes

production

Platform architecture for industrial flexibility, smart energy operations, telemetry ingestion, market integration, scheduling, billing, REMIT reporting, analytics, optimization, and Kubernetes-based deployment. The system connects industrial protocols, market APIs, time-series services, and operational workflows into a production platform.

Context: Industrial energy · real-time telemetry · flexibility operations
Role: Principal architecture · platform design · production deployment guidance

Multi-protocol

MQTT, OPC-UA, Modbus, DLMS, REST

Kubernetes HA

Resilient deployment architecture

Hundreds DB tx/sec

Live PostgreSQL platform workload

ML platform for energy forecasting

Python · model lifecycle · time-series · prediction workloads

production

Specification and architecture for a Python platform that registers models, manages model instances and parameters, schedules training and prediction, stores time-series outputs, monitors production execution, and supports forecasting workloads across 200+ model instances and roughly 2M predictions per year.

Context: Production ML · 200+ model-instance scale · ~2M predictions/year
Role: Platform architect · model lifecycle and Python framework design

200+

Model-instance scale

~2M/year

Forecast values produced

Production RL trading system

Python · reinforcement learning · Kraken spot and futures

production

Live SOL/USD spot and futures trading system on Kraken, designed and implemented end to end in Python, with data ingestion, Level-2 order-book reconstruction, CRC32 data verification, walk-forward validation, regime switching, independent risk layers, server-side stops, dead-man switch, hot model reloads, and monthly retrain gating.

Context: Live on Kraken · spot and futures · real capital
Role: System architecture · Python ML implementation · production operation

18 months

Development duration

Live spot + futures

Production status

Adaptive AI orchestration system

Python · multi-LLM · persistent memory · local inference

active

Python-based AI automation platform for multi-step workflows, with cognitive mode selection, episodic memory in PostgreSQL and Qdrant, multi-LLM execution, workflow engine, tool registry, and deployment paths for local, Docker, or Kubernetes environments. Learned patterns can move from expensive reasoning to lower-cost local execution.

Context: Active use · AI automation · memory-driven workflows
Role: Architecture · Python implementation · memory and orchestration design

PostgreSQL + Qdrant

Persistent episodic memory

Cost-aware routing

Premium, strategic, and local execution

Architecture

How complex systems become reliable in production.

Across insurance, energy, trading, ML, and AI automation, the same pattern appears: understand the domain rules, structure the data and computation, then make the system observable, testable, deployable, and usable by the team.

Domain logic

Business rules
Data contracts
Edge cases
Ownership boundaries

Engineering core

Python
C# · PL/SQL
Oracle · PostgreSQL · Azure · Kubernetes
Event pipelines · time-series

Production operation

Validation
Monitoring
Deployment gates
Team enablement

Operating outcomes

What the work changes in production.

The outcome is continuity, correctness, automation, risk control, live operation, and teams that can keep improving the system.

01
Core insurance continuity
Continuity and evolution for an Amarta/FIS life administration system supporting 100K+ life policies across traditional, unit-linked, annuity, endowment, and risk products. Over many years, the platform supported introduction of new and innovative insurance products while daily servicing, premiums, claims, commissions, reinsurance, underwriting, data warehouse workloads, and operations remained stable.
100K+ policies
02
Regulatory framework foundation
C# framework foundation for Solvency II / IFRS 17 applications, separated from actuarial calculations implemented by actuarial teams.
S2 / IFRS 17
03
Energy trading and risk automation
Python data warehouse and calculation engine made gas, futures, supply P&L, hedging, mark-to-market, risk limits, controls, and notifications repeatable in production.
P&L + risk controls
04
Production ML forecasting
Model lifecycle architecture supports registration, training, prediction, parameters, time-series outputs, monitoring, alerting, and roughly 2M forecasts per year.
~2M/year
05
Risk-controlled live execution
Live Kraken spot and futures operation built around data verification, walk-forward validation, independent risk controls, deployment gates, dead-man safety, hot reloads, retraining, and monitoring.
Live on Kraken
06
Adaptive AI automation
Memory-driven AI orchestration routes tasks between premium LLMs, strategic planning, and local execution based on history, success rates, failures, and complexity.
Cost-aware routing

External perspective

What senior leaders say

Short references from insurance, finance, energy, and regulated-system environments.

Edvard Šimec

Director of Information Technology and Business Operations · Generali Slovenija

Insurance · life insurance · back-office systems

Dušan and I have collaborated very successfully for many years on the maintenance of one of our key back-office systems. He stands out for his deep knowledge of insurance, particularly life insurance, combined with a broad command of the full IT spectrum: from architecture and business analysis through to development. His expertise and reliability contribute significantly to the stability and continued evolution of our solutions.

Serge Runge

Enabling Manager Flexibilitätsvermarktung · enercity

Flexibility marketing · direct marketing · energy optimization

What truly sets Dušan apart is his deep understanding of how software for the energy domain is evolving. Asset owners, operators and traders expect white box solutions, technology-driven and designed to empower users to contribute more actively. Dušan helped us establish a software architecture that makes critical components in asset management and multi-market optimization easy to configure and extend - delivering stable ground for key use cases and openness for some degree of co-development.

Dr. Darko Medved

Founder / Managing Director, JMD Consulting

Insurance · actuarial · regulatory systems

Dušan understands the depth of modern insurance systems. He is able to architect complex solutions where business rules, actuarial logic, regulatory requirements, data, and long-term maintainability all have to work together. His work reflects a strong understanding of both the insurance domain and production technology.

Jožica Palčič

President of the Management Board · Sava Infond

Financial services · governance · operational control

Dušan is not only a strong technical architect; he understands the management perspective behind critical systems. He can structure technology in a way that supports business continuity, operational control, and confident decision-making.

Nigel Gardner

Business Development Executive · Charles Taylor InsureTech

Life insurance technology · policy administration · insurance platforms

I have worked with Dušan in the life insurance technology space and value his combination of domain knowledge, technical depth, and delivery discipline. He understands how complex policy administration and insurance platforms need to operate in real client environments.

About the operator

Dušan Krovinović

Former CIO · enterprise architect · principal engineer · 30 years

30 years · former CIO · Python/SQL systems · insurance, energy, trading, AI · production-first

Former CIO, enterprise architect, PMO lead, product manager, and lead developer with 30 years across systems where failure is expensive: life insurance administration, regulatory frameworks, Python data platforms, industrial IoT, production ML, live market execution, and AI-assisted engineering workflows.

Focus: turning domain complexity into systems that keep producing correct results in production, and leaving teams with workflows they can keep using.

Need senior ownership of a system where correctness, risk, and production continuity matter? Let's discuss.

Principal architecture · Python data platforms · regulated systems · production AI · live trading infrastructure.

Start a conversation LinkedIn

dkrovinovic@elysian-systems.com

Response within 24 hours (EU timezone).