Askona — Enterprise Architecture and AI Transformation

Role: architecture/engineering leadership in CTO scope | Span: 20+ teams, 40+ systems

$680M+: revenue
15M: customers
9000+: employees
40+: services

Led a controlled digital and AI transformation in a high-stakes enterprise environment to speed up delivery, improve reliability, and institutionalize architecture standards without breaking business continuity. A core part of the work was a multi-party integration layer — a shared fabric for many independent owners (teams, hybrid cloud and on-prem, partner footprints) so integration seams stayed explicit and contracts did not drift.

Role & authority

Position: Enterprise Architect (working within the CTO's organization)
Reported to: the CTO/CIO
Scope: Enterprise architecture and AI-focused transformation; ran the service ecosystem as a product (own roadmap, standards, monitoring); coordinated 20+ distributed teams across 40+ systems; drove AI adoption top-down across business functions — up to a dedicated AI department.
Leadership model: Matrix leadership across 20+ teams — architectural authority, hiring calls and governance; the way leadership works at enterprise scale. Organization: $680M revenue, 15M customers, 9000+ employees (largest mattress & furniture retailer in Eastern Europe).

Recommendations

From Askona colleagues (LinkedIn)

"Grigoriy consistently demonstrated exceptional technical leadership and deep expertise as an Enterprise Architect. He was one of the key drivers behind the adoption of event-driven architecture and microservices across Askona's IT landscape — pragmatic, collaborative, and always focused on the right solution for both business and technology."

Aleksander Zinger · Director-level IT Business Partner, Askona

"I highly recommend him as an exceptionally competent professional in Enterprise Architecture, Solution Architecture, AI-driven innovation, and technical leadership. He led the company's strategic AI initiative, driving the adoption of AI tools and practices across the organization."

Ilya Chernov · Head of PMO IT, Askona

Situation

• Context: large enterprise (15M customers, 9000+ employees, $680M+ revenue).
• Architecture landscape: hybrid cloud + on-prem, 40+ systems, high integration complexity.
• Business pressure: faster change delivery with no compromise in stability.
• Critical risk: conservative teams, low business engagement, and over-ambitious change scope.
• Marketplace-style framing: delivery depended on a multi-party integration layer—event buses, APIs, and governance as the coordination fabric across independently owned services and teams, not “just an AI project.”

Task

• Increase release cadence and time-to-market without production instability.
• Reduce delivery friction across teams and systems.
• Modernize legacy under controlled governance.
• Move AI from demos into real operating workflows.
• Keep integration seams predictable with many system owners: shared exchange rules and boundaries so changes in one part of the landscape do not break adjacent systems.

Action

• Coordinated transformation across 20+ distributed teams under one architecture frame.
• Built event-driven integration (Kafka/RabbitMQ/API) as the backbone between independent teams and services—explicit contracts at ownership boundaries, faster launches, less integration friction. The backbone handled millions of events/day across 40+ systems.
• Implemented platform governance and high-availability standards.
• Shifted execution to phased rollout instead of a big-bang rewrite.
• Led AI seminars, workshops, and 1:1 enablement with teams and managers.
• Introduced practical RAG, AI-assisted coding, and AI use cases for delivery and business decisions.

Skills applied

• Enterprise architecture and coordination of 20+ distributed teams across 40+ systems.
• Multi-party integration layer: aligning many owners, contracts, and operating models in one production footprint.
• Event-driven integration (Kafka, RabbitMQ, APIs) to reduce integration friction.
• Platform governance, high availability, and phased legacy modernization without a “big bang.”
• AI enablement: RAG, AI-assisted / vibe coding, and applied AI patterns within security constraints.
• AI leadership: seminars, workshops, and 1:1 guidance for teams and management.
• Transformation management with a strong focus on business continuity and production stability.

Result

Operational impact

stable gains visible in first 2 years

Cultural impact

by year 3-4 teams defended new standards

Architecture impact

event-driven principles became company standard

Org impact

AI practices became the foundation of a dedicated AI function

Strategic outcome

Release cadence and time-to-market accelerated, platform reliability increased, cross-team friction decreased, and bottleneck visibility improved. Integration seams between parts of the landscape became more predictable—fewer one-off manual alignments for every change. Under my leadership, AI became the primary operating pattern across all levels — from management decisions to engineering execution.

Public Askona business context from open sources: 2023 revenue under RAS was RUB 49.12B (vs RUB 40.94B in 2022), and for key legal entity "TD Askona" open 2024 reports mention RUB 41B revenue and RUB 1.2B net profit. These figures describe company-level performance and are not a direct attribution to this transformation alone.

Org-wide AI adoption (beyond engineering)

AI did not stay inside engineering — it spread across most of the company's business functions and became a normal way of working well beyond IT.

• Top-down demand-pull, not a push at engineers. I drove adoption from the top — into the business and executive layer first (the CEO and one or two levels below), so demand formed at the decision-making layer. Management then began requesting AI capability themselves and asking for it as a standing practice; adoption became organic pull rather than a mandated rollout.
• Cross-functional reach. Use spread beyond IT — logistics, production, flagship retail, 1C, finance and budgeting. Non-technical staff built their own small tools for concrete operational problems (for example, a calculator for how many cartons fit in a truck), a marker of how deeply it penetrated.
• Three adoption tracks. (1) AI-assisted requirements validation, so specs were clarified quickly and rigorously; (2) automated approvals, including an agent-to-agent workflow that reconciled specifications; (3) manager-led enablement — teaching managers and their people to build their own solution (a compute utility, microservice, or dashboard) when a real business need had no engineering resource, budget, or timeline.
• Curve and endpoint. Roughly 6–8 months from the first "let's use AI" proposal to the practice being institutionalized as a dedicated AI department with its own head — the clearest proof of stickiness: it outlived the individual push and became part of the company's structure.

Practical examples

• AI bot over data lake: business questions in natural language were translated into SQL, with both human-readable and machine-readable outputs. Analyst ad-hoc turnaround dropped from days to seconds, in regular use; read-only access and schema-name validation as guardrails.
• Next best offer: I personally designed and trained an ML model on historical order data for recommendations in cart and email scenarios.
• Order lifecycle timestamp tracker: implemented end-to-end event tracking to surface bottlenecks and delay zones in order processing.
• Georesolving proxy microservice: as part of AI-assisted coding training sessions, built a service that converts coordinates to address and address to coordinates for integration needs.
• AI-agent demo stand: assembled a cross-organization setup to demonstrate collaborative specification alignment between agents and reduce manual coordination friction.
• Training scale: delivered dozens of sessions on engineering practices, management, and applied AI for audiences of 100-200 participants.

Personal contribution

• Institutionalized event-driven integration and phased transformation without a big-bang rewrite—in practice the multi-party integration layer for 40+ systems and 20+ teams with separate owners.
• Personally delivered the key AI/ML assets: data-lake AI bot, next best offer ML model, timestamp tracker, georesolving proxy service, and AI-agent demo stand.
• Scaled adoption through recurring enablement for teams and managers, turning AI-assisted coding into an operating practice rather than a one-off demo.

Case Materials

• youtube: YouTube channel
• distributed asynchronous systems course: YouTube playlist
• blog (ru): dobryakov.com/blog

Grigoriy Dobryakov — AI-Driven Head of Engineering

$680M+ revenue 40+ systems, 20+ teams

Connect as MCP Email me