ML / Ops / AI Engineer

На данный момент Lead ML/AI engineer.

Мои компетенции и опыт

# Backend-focused Senior ML/Lead Engineer

Backend-focused Senior ML/Lead Engineer building production Python systems with FastAPI, async services, external integrations, and cloud-native infrastructure. Delivered backend platforms from vague product ideas to shipped features: API design, ETL/data services, high-load distributed systems, CI/CD, and testable engineering workflows. Hands-on experience with LLM integrations, text-to-SQL, document processing, and data-driven products; also automated real estate workflows for multiple clients through process mapping and platform delivery.

## Skills

### Backend

Python, FastAPI, asyncio, SQL, PostgreSQL, gRPC, Celery, RabbitMQ

### API & Integrations

API Design, External Data Integrations, REST APIs, OpenAPI, Text-to-SQL

### Testing & Code Quality

pytest, TDD, Code Review, Pyright (strict), GitLab CI, CI/CD

### Data & Storage

ClickHouse, PostgreSQL, AirFlow, Pandas, Polars, PySpark, MLflow

### LLM & AI

LLM Integration, Prompt Engineering, Document Processing, Embeddings, Vector Search

### Infrastructure

Docker, Kubernetes, Terraform, Azure ML, AzureML Pipelines, Monitoring

## Experience

### Lead Engineer (ML, AI, Real estate) — First Line Software

Remote | 09/ нужен доступ к резюме now

нужен доступ к резюме

- Built backend and data platform components for a real-estate analytics product, automating client processes across multiple engagements and translating vague requirements into production deliverables.

- Designed APIs and integrations with external data sources; implemented Azure Blob-based storage, ETL flows, CI/CD, and model-serving REST endpoints with zero-downtime updates.

- Created an AI chatbot with text-to-SQL for querying financial and operational indicators, plus an LLM-based document-processing service for reports, letters, and document abstracts.

- Built a production-parity evaluation harness for a tool-calling LLM agent using in-process FastAPI, ephemeral PostGIS + DuckDB, Testcontainers, and ASGI transport for deterministic regression testing without external infra.

- Improved backend performance and engineering reliability through async execution design, storage-agnostic MLflow/in-memory recording, observability, alerting, and mentoring of junior engineers.

### ML-Ops Engineer — Quantum Brains

Remote | 08/ нужен доступ к резюме /2024

нужен доступ к резюме

- Architected and productionized a distributed simulation platform from a local engine into a cloud-native service stack using RabbitMQ and HashiCorp Nomad.

- Designed APIs and translated internal customer requirements into roadmap items, tickets, and delivery plans for features that became mission-critical for multiple ML teams.

- Optimized performance and reliability under heavy concurrent load, reducing memory usage and improving resource allocation for 300+ node backtests.

- Led engineering execution for junior developers with code reviews and CI/CD quality gates using pyright, ruff, GitLab CI, and pytest.

- Implemented feature-store integrations and a custom cloud autoscaler extending standard orchestration capabilities.

### ML Engineer — Workestra

Remote | 07/ нужен доступ к резюме /2023

нужен доступ к резюме

- Owned end-to-end backend data pipelines for a retail forecasting product, integrating heterogeneous client data sources and ensuring reliable availability for ML and analytics.

- Improved platform stability by resolving production outages and strengthening monitoring and alerting for customer-facing optimization services.

- Modernized infrastructure by migrating file storage from local VM to S3 and moving analytical datasets from Postgres to S3 Parquet to reduce database load.

- Replaced Celery Beat orchestration with Airflow to improve observability, maintainability, and scalability of scheduled workflows.

- Developed forecasting models that improved R² by нужен доступ к резюме , supporting a product outcome tied to company fundraising.

### Machine Learning Researcher — Schlumberger Moscow Research

Moscow | 08/ нужен доступ к резюме /2022

нужен доступ к резюме

- Developed a two-phase petroleum reservoir model compatible with OpenAI Gym-style environments.

- Applied reinforcement learning algorithms for oil well control and improved discounted revenue by 20% versus a no-control policy on a test scenario.

- Built integration between the RL control solution and a commercial reservoir simulation engine.

### Machine Learning Intern — Acronis

Moscow | 09/ нужен доступ к резюме /2020

нужен доступ к резюме

- Developed time-series forecasting models for server response latency.

- Implemented data loading from Prometheus.

- Added forecasting as a feature for anomaly detection.

## Education

### Master of Computer Science — Moscow Institute of Physics and Technology

нужен доступ к резюме

Coursework: Machine Learning, Data Science, Computational Mathematics

### Bachelor of Computer Science — Moscow Institute of Physics and Technology

нужен доступ к резюме

Coursework: Probability Theory and Statistics, Linear Algebra, Computer Science, Data Structures, Algorithms

## Publications

### Autonomous Reservoir Management with Deep Reinforcement Learning

European Association of Geo-scientists and Engineers | Novosibirsk | 08/2022

нужен доступ к резюме

Application of Reinforcement Learning to reservoir control. Presented a prototype improving revenue by 20% on the test case.

### Autonomous reservoir control with Deep Reinforcement Learning

European Association of Geo-scientists and Engineers | Moscow | 11/2022