Igor Gorbenko

Compliance App is an enterprise compliance and audit automation platform designed for organizations operating under complex regulatory requirements, including PCI DSS.

I serve as CTO and Software Architect and own the technical architecture of the platform across application development, infrastructure, security, data, and AI capabilities. I designed a multi-tenant architecture supporting both SaaS and on-premise enterprise deployments. The platform includes audit management, evidence collection and validation, asset and risk management, vulnerability management, workflow automation, reporting, and integrations with external enterprise systems.

One of the key areas I designed is the platform's AI layer. It includes a private AI assistant based on LLM and RAG technologies that can work with regulatory documentation, audit evidence, internal policies, and platform data. The architecture is designed to support private and on-premise deployments where sensitive customer data cannot be transferred to public AI services.

My role combines CTO responsibilities, solution architecture, AI architecture, engineering management, and direct collaboration with customers on complex technical and regulatory requirements.

I designed and led the implementation of a large-scale Data Lakehouse for an AdTech platform processing more than 2 PB of data and handling workloads associated with a platform operating at up to 300K requests per second.

The architecture was built on Amazon S3 and Apache Iceberg, with Apache Spark as the primary distributed processing engine. The platform processed up to 10 TB of new data per day and supported both batch and near-real-time ingestion patterns.

I designed two major ingestion paths. The batch layer handled high-volume historical processing, transformation, and aggregation workloads, while the streaming layer consumed continuously generated events and made fresh data available for analytics and downstream services.

Apache Iceberg provided a scalable table layer with schema evolution, partition management, and transactional data operations, while Spark was used for complex transformations and large-scale distributed computation.

As Head of Data and Software Architect, I was responsible not only for the technical architecture but also for the overall data strategy, engineering standards, infrastructure decisions, and integration of the Lakehouse with analytical and operational systems.

I led the architecture and engineering of a large-scale real-time recommendation platform for Tango, a global live video streaming platform.

The system continuously analyzed user behavior and content signals to select and rank the most relevant live content for each user. The architecture was designed for low-latency online inference and high-throughput event processing.

The platform was built primarily on Google Cloud and used Bigtable, BigQuery, Dataflow, Redis, and Python-based services for feature processing, candidate generation, ranking, and real-time recommendation delivery.

I owned the architecture and engineering delivery across the data, machine learning, and cloud layers and led the ML Engineering organization responsible for the platform.

In addition to the recommendation platform, my team developed real-time anti-fraud and video content analysis systems. The recommendation initiatives contributed materially to Tango's revenue growth, while the anti-fraud platform reduced false-positive rates by more than 80%.

My responsibilities included architecture, engineering leadership, MLOps practices, production reliability, team development, and collaboration with product and business stakeholders.

I led the development of a data transformation framework for converting heterogeneous healthcare datasets into the OMOP Common Data Model (CDM).

The main engineering challenge was supporting datasets originating from multiple technologies and storage systems, including Amazon S3, Google Cloud Storage, Hadoop HDFS, PostgreSQL, Amazon Redshift, and other relational and distributed data platforms.

We developed a reusable framework that automated the preparation and execution of large-scale ETL pipelines using Python, Apache Spark, and SQL.

The framework reduced the amount of manual work required to onboard new datasets and provided standardized orchestration of transformation stages. It also supported automated post-processing steps such as data validation, unit tests, statistics generation, and quality reports.

As Technical Lead, I designed core framework components, developed Python services, reviewed code, helped define ETL architecture, and participated in running and troubleshooting production data pipelines.