● irina@irinasafonoff ~ portfolio
whoami

Irina Safonova

Lead Software Engineer & Solution Architect — Data Platforms, Enterprise Integrations & Practical AI

I design and build data-intensive systems for complex business operations.

For more than 25 years, I have been turning operational complexity into production software: enterprise platforms, financial workflows, data products and AI-enabled solutions.

I combine solution architecture with hands-on engineering, moving between system design, data modelling, implementation and delivery as the product requires.

My focus today is practical AI: using AI where it adds measurable value while keeping core business logic deterministic, reliable and auditable.

See what I build Hiring? Get in touch Discuss a business problem Download CV (PDF)
Ships production systems, not prototypes Data architecture first Bridges engineering ↔ business
LinkedIn
Portrait of Irina Safonova Helsinki, Finland
In software since
2000
Master's degrees
2
Team leadership
6 engineers
International scope
Systems across 3 countries
How I can help

Where I add the most value

Two kinds of people end up on this page — someone building a team, and someone with a system that needs fixing. Here's what each gets from me.

If you're hiring

Solution architecture, start to finish

I design data and AI platforms from scratch, or untangle ones that grew faster than their architecture — data model, pipelines, APIs, security.

The bridge between business and engineering

I translate messy, half-formed requirements into technical plans, backlog items and something a team can actually build against.

Delivery, not just design

I work collaboratively with engineers, validate architecture against implementation constraints, and stay hands-on where the team needs me. I can lead technical design, contribute directly to the codebase, or combine both.

If you need a consultant

I help companies modernize data-intensive systems, automate complex operational workflows and introduce AI without compromising reliability, security or auditability.

Architecture and system review

For systems that have become difficult to change, expensive to maintain or dependent on manual work. I review the architecture, data flows, integrations and operational bottlenecks, then define what should be fixed first.

Deliverables: system map, risk assessment, recommended target architecture and prioritized implementation roadmap.

Data and reconciliation discovery

For teams that compare records across billing, finance, ERP or operational systems and spend too much time investigating mismatches manually. I map the sources, matching rules and exception workflow, then design a reliable solution.

Deliverables: source and process map, reconciliation model, exception workflow, MVP architecture and implementation plan.

Practical AI assessment

For companies considering AI but unsure where it will create measurable value. I identify realistic use cases, assess data readiness and define where deterministic software should end and AI-assisted workflows should begin.

Deliverables: prioritized use cases, data and risk assessment, build-versus-buy recommendation and technical plan.

Discuss a business problemNot sure which one fits? Just write. Happy to figure it out together.

Core expertise

Toolkit

AI & Integrations
Applied ML — academic & project experience
LLM integrations (OpenAI, Azure OpenAI)
AI-assisted & agentic workflows
RAG-style data access
Languages & Data
SQL Server / T-SQL (25 yrs)
Python (10 yrs)
Performance tuning
Architecture
Data-intensive & multi-tenant systems
REST APIs & enterprise integration
Backend automation & event-driven processing
Platform & Infra
Azure project experience, including Azure OpenAI & Event Hub
AWS project experience
Docker
Terraform working knowledge
Delivery
Agile / Scrum (CSM)
Implementation planning & delivery
Consulting & stakeholder communication
Leadership
Solution architecture
Tech lead / team lead
Cross-functional bridging
Selected work — 2000 to today

25+ years of production systems
the timeline, and the ones worth reliving

Scan the timeline below for the quick version, or read on for the reconstructed screens and stories behind each role. Client and tenant data is redacted where confidentiality applies.

2026 · contractLead Solution ArchitectEarly-stage B2B SaaS (confidential)
2024 – 2026Technical IT Consultant → Senior Software EngineerGood Sign
2017 – 2024 · 7 yrsBusiness & IT ConsultantSelf-Employed
2012 – 2015 · 3 yrsSAP/ERP Migration ConsultantStroytechkom
2007 – 2012 · 5 yrsSoftware DeveloperAl-Arqam School, Qatar
2000 – 2008 · 8 yrsTeam Lead / Software EngineerGiprosvyaz
2026 · Contract Lead Solution Architect

Secure multi-tenant data and AI platform

An anonymized early-stage B2B SaaS platform that brings data from multiple external systems into a consistent internal model and supports reliable, auditable processing across isolated client environments.

Deterministic services remain responsible for trusted business logic. AI-assisted components are isolated and used only for controlled interpretation, explanation and investigation support over authorized data.

My role: target architecture, data model, ingestion workflows, tenant isolation, AI integration boundaries, deployment structure and implementation planning — with hands-on contribution where required.

Temporary contract; product details remain confidential, while the architectural approach is described here.
Read the architecture case
app.confidential-platform.com/operations
Data operations
Tenant isolatedIS
01 Sources
3 connectorsAPI, SFTP file and structured upload
02 Ingestion
12,486 recordsValidated, typed and deduplicated
03 Normalize
Shared domain modelSource records mapped to consistent entities
04 Match
Deterministic engineMatches, mismatches and rule outcomes
05 Review
AI-assisted analysisEvidence-grounded exception explanations
Data sources Last 24 hours
ERP ledger
Tenant-scoped connector
8,220 records
Completed
Payment provider
CSV schema v4
3,914 records
Processing
Billing export
Validation required
352 records
7 exceptions
Authenticated tenantScoped storageAudit trailEncrypted secrets
Exception review Evidence attached
Reference mismatch
4 records could not be matched automatically
Amount mismatch
3 records differ across source systems
AI-assisted explanation
A recent source-format change may explain the validation exceptions. The explanation is evidence-grounded and cannot alter trusted system results.
Conceptual reconstruction of an anonymized multi-tenant data platform. It shows source ingestion, normalization, deterministic processing, exception review and a constrained AI explanation layer. All data is fictional.
2024 – 2026 · Technical IT Consultant → Senior Software Engineer

Good Sign — SaaS billing & pricing

Worked on a SQL-centric, data-intensive SaaS platform supporting billing operations for hundreds of client organizations. Customized stored procedures, billing logic and data workflows for client-specific requirements; designed and implemented SQL-based solutions for complex billing, pricing and integration scenarios; investigated production data issues and improved performance and maintainability. Contributed to AI-assisted workflows over structured enterprise data and took on Scrum Master responsibilities when needed.

Client data and implementation details are confidential; responsibilities and technical scope are summarized accurately.
2017 – 2024 · 7 yrs · Business & IT Consultant

Reporting & workflow automation

Independent consulting for enterprise clients — legacy modernization, data-flow redesign, and reporting automation. The unglamorous plumbing that quietly saves a finance team its Friday afternoons.

2012 – 2015 · 3 yrs · SAP/ERP Migration Consultant

Factory SAP-to-local-ERP migration

Led the migration of a factory's core operations from SAP to a local ERP alternative for Stroytechkom — data migration, business-process remapping, and a cutover planned around not stopping the production line.

2007 – 2012 · 5 yrs · Software Developer

Student assessment system

A desktop application built for Al-Arqam International School, Qatar: moved grade entry, term averages and report cards off paper and into one system, cutting out the manual paperwork teachers used to fill in by hand every term — report work that took 3–4 days dropped to about half an hour. No dashboards, no charts — a grid, a database, and evenings teachers got back.

2000 – 2008 · 8 yrs · Team Lead / Software Engineer

ERP for the Giprosvyaz institute

Enterprise resource planning for a telecom design institute — contracts, project stages, and deadlines for fiber-optic and exchange projects across the region. Built the integration with the institute's CRM and automated the manual postings the accounting team did by hand, cutting their processing time by 70%; reports that used to take an hour to compile now generate straight from the ERP in about 10 minutes. A relational database behind, a six-person team I led through eight years of releases.

Giprosvyaz ERP — [Contracts]_□×
FileEditReferencesDocumentsReportsToolsWindow?
Contract No.CustomerDesign ObjectAmount, RUBStageDue Date
214-05Lensvyaz JSCFiber line St. Petersburg–Vyborg, working docs1 840 200WD30.06.2005
198-04Novgorod PSTNATS-63 reconstruction, station facilities976 500DP15.03.2005
187-04NW Telecom JSCGatchina district network, line facilities2 310 000WD30.09.2005
165-03ElektrosvyazCable ducting, Petrozavodsk418 300WDdelivered
AddEditDeleteReport…
Records: 1,247User: I. Safonova14.03.2005
Reconstruction — contract registry of the ERP built for Giprosvyaz institute, 2000–08
Education & credentials

Foundations

M.Sc. Software Engineering & Digital Transformation

LUT University, Finland · 2021–2023

M.Sc. Communication Networks & IT

State University of Telecommunications · 2001–2007

Automated Control Systems for Aircraft — B.Sc.

D.F. Ustinov Baltic State Technical University "Voenmeh"

Automated Database Management Systems

Physics & Mechanics Engineering College
Certified ScrumMaster (CSM) AWS Cloud Quest Practitioner IoT IBM Cloud Computing

Working stack: Python, SQL Server / T-SQL, Kafka, Azure Event Hub, Docker, Terraform, Azure, AWS, OpenAI and Azure OpenAI.

Also served as Scrum Master (Oppi AI, LUT, 2022), guiding Agile delivery for an internal team, and as Adjunct Lecturer (LUT University, 2023), teaching software engineering courses.

Get in touch

Building a data or AI platform? Let's talk architecture.

Based in Helsinki, working with teams anywhere. Consulting on data architecture, AI integrations, and making legacy systems earn their keep.

LinkedIn Email

References available on request. · Download CV (PDF)

Ask my portfolio assistant
Answers from the information on this site