Designed and implemented from scratch:Collaborative and software engineering tools for the teamCI/CD, including homegrown IaC (Terrafrom, Helm) continuous deploymentInitial product prototype (chatbots, dialog engine)Cloud infrastructure (GCP, AWS, Azure and specialized like Lambdalabs and HyperStack), both managed and unmanaged Kubernetes for GPU workloadsB2C and B2B authentication: Identity platform, OAuth2, JWTAsset management, incident management automated processesHR, contracts and purchasing processes automationB2B API monetizationMLOps toolingData warehouse / Data lake / BI toolsTech stack of the projects includes:GCP: GKE, Dialogflow, Firebase, Apigee, Vertex AI, Cloud Run, Pub/Sub, Cloud Build, IAP, Bigquery, ApigeeAWS: EKSAzure: AKSAtlassian: Jira, Confluence, Bitbucket, Gitlab, OpsgenieClearML, Stripe, Telegram/Whatsapp/FB messenger bot API, MailchimpFastAPI, Terraform, Flywaydb, PostgreSQL, Redis, ChatGPT API, OpenTelemetry, Kong API gateway

Building managed Hadoop and Spark infrastructure for Yandex.Cloud Data Proc product.Developing and maintaining of Yandex.Cloud Managed service for Apache Kafka.Among other things I have implemented:- Hadoop cluster graceful decommission support- Hadoop cluster autoscaling feature- Custom authorizer module for Apache Kafka- Hadoop cluster cloud logging and monitoring- Yandex.Cloud integration plugin in InfluxDB Telegraf- Yandex.Cloud integration for Apache Airflow- Automated certificate renewal for Kafka clustersTook part in developing things such as- CI/CD infrastructure for Managed Hadoop/Kafka products ( TeamCity, Jenkins, Hashicorp's Packer etc) - Yandex.Cloud Data Proc integration in Hashicorp's TerraformTech stack of the projects includes:Python, Go, Java, SaltStack, Jinja2, gRPC, Postgres, Clickhouse, Kubernetes, cloud solutions

Developing user activity data aggregation platform to be used inhouse for ad targetting and recommendation systems. Building and exporting user behavior profiles with various machine learning approaches, combining heterogeneous data sources up to voice recognition models.Creating and optimizing batch and realtime data pipelines with different data engineering frameworks.Collecting and reporting business intelligence metrics.Managing ML ops processes:- Monitorings, automatic evaluation of models- Evaluating workflows for experiments and ad-hoc tasks- Duty roster management- DMP worklows ( ingesting data from external sources )Tech stack includes:Word2vec, Tensorflow, Keras, CatBoost, Vowpal Wabbit, Docker, Spotify's Luigi, Apache Storm, Pandas, Scikit-learn, Jupyter, proprietary analogue of Hadoop/Hive, proprietary analogue of Apache NiFi, Python, Java, Ubuntu

Building the infrastructure for user behavior profiling

- Developing malware and unwanted content detection system that is used to process 20m+ webpages and thousands of binaries daily.- Reverse engineering of exploits met in the wild. - Detectors quality control management.- Ad-hoc forensics and incident management.- Doing tech talks and publications.Among other projects I have accomplished I can mention the dynamic detectors of Java and .NET exploits based on modified Java and .NET runtimes, which includes machine learning detection approach as well.And one of my pet projects at the time: application firewall and malicious activity detection system for Android, based on automatic binary-level instrumentation of third-party apps https://github.com/peter-volkov/Patcher5Tech stack includes:IDA, ImmunityDebugger, WinDbg, Fiddler, WireShark, HyperV, VMware Server, Cuckoo sandbox, Python, JavaScript, PHP, Swagger2 / OpenAPI, Headless browsers, Windows, Powershell, Android