# Nazarii Melnychuk – Lead Data & Applied AI Engineer

> Lead Data Engineer and Applied AI Engineer specializing in data-platform modernization, lakehouse architecture, and production agentic systems. Data Lead for a team of 6 engineers (grown from 2 through hands-on hiring) and lead of a 9-person cross-functional delivery team. A generalist who works primarily in Python and reads Scala, Go, and TypeScript, focused on architecture, performance, business correctness, and design elegance.

This is the Markdown representation of the CV published at https://nmeln.com/cv — the same document browsers see as HTML. The CV is a single page and is reproduced here in full, so no other pages need to be fetched.

## Contact & links

- Website: https://nmeln.com
- LinkedIn: https://www.linkedin.com/in/nazarii-melnychuk-84581111b
- GitHub: https://github.com/nmeln
- Contacts (Linktree): https://linktr.ee/nmeln
- JSON Resume: https://nmeln.com/resume.json
- This CV as Markdown: https://nmeln.com/cv.md

## Major Points

- Lead Data Engineer and Applied AI Engineer specializing in data-platform modernization, lakehouse architecture, and production agentic (LLM) systems
- 10 years of experience across distributed systems, large-scale data pipelines, and applied AI
- Directly managed 6 data engineers, including hiring, onboarding, regular 1:1s, development planning, performance feedback, and workload allocation; grew the team from 2 to 6
- Also lead a 9-person cross-functional delivery team, keeping technical direction aligned with business goals
- Experienced in roadmap ownership, capacity and dependency management, automated data validation, root-cause analysis, privacy-sensitive processing, and production AI systems
- Strong technical partner to ML engineers and data scientists, with a current focus on experiment tracking, data lineage/registries, data quality, and synthetic data
- Generalist engineer: primary language **Python**; reads Scala, Go, and TypeScript
- Driven by architecture, performance, business correctness, and design elegance

## Specialized in

- Data-platform modernization and **lakehouse** architecture (**Databricks**, **Delta Lake**, medallion pipelines)
- Large-scale batch and streaming data processing with **Apache Spark**
- Production applied-AI and agentic systems: **LLM** services, agent frameworks, **MCP** tools, embeddings, and vector search
- Scalable pipelines and services in **Python** on **AWS**, with a focus on performance and business correctness

## Leadership and mentoring

- Serve as Data Lead for 6 data engineers, growing the team from 2 to 6 across sourcing, interviewing, technical assessment, hiring, onboarding, and development
- Lead a 9-person cross-functional delivery team (including 4 data engineers), coordinating technical direction, priorities, dependencies, and execution
- Own a multi-quarter data-platform roadmap, translating business goals into architecture decisions, implementation plans, and launch criteria
- Navigate complex organizational change and shifting priorities while preserving delivery reliability and measurable business outcomes
- Technical interviewer; previously ran Scala bootcamp lectures and internal workshops to upskill mid-to-senior engineers

## Industries

- Cloud communications (WhatsApp, RCS, SMS)
- Marketing/Adtech (user engagement data, ad analytics)
- Healthcare data processing and HIPAA compliance
- IoT monitoring for food/beverage companies
- Energy sector digital transformation

## Achievements

### Applied AI & automation

- Built and launched an LLM-powered natural-language data discovery service from zero to production (FastAPI, agent frameworks, MCP, embeddings, and vector search).
- Designed and productionized a Claude-powered autonomous incident-triage agent with permission-gated, read-only investigations and an auditable activity trail.
- Automated data-team and sprint/planning processes as reusable agent skills, codifying runbooks into workflows with confirmation steps and invariant checks.

### Communications project

- Developed in-house services replacing costly external APIs, generating $100k-$250k customer savings
- Optimized message routing and billing pipeline during Kinesis-to-Kafka migration
- Achieved improved throughput (830+ MPS) by eliminating processing redundancies
- Implemented comprehensive cross-stream validation to ensure data consistency between source systems during this critical transition

### Healthcare project

- Transformed legacy Python batch jobs into scalable Apache Spark workflows, successfully migrating complex healthcare data processing logic
- Contributed significant system architecture decisions across multiple projects to build production-grade ETL pipelines handling diverse healthcare data formats

## Skills

- Languages: Python (primary), Scala, Go, TypeScript
- Data & Lakehouse: Databricks, Delta Lake, Apache Spark, Apache Kafka, Airflow
- Applied AI: LLMs, Claude / Claude Code, MCP, FastMCP, Embeddings, Vector search, FastAPI
- Agent SDKs: OpenAI Agents SDK, Claude Agent SDK, Inngest AgentKit
- Data Warehousing & Analytics: Amazon Redshift, BigQuery, Amazon Athena, ElasticSearch, pandas
- Databases: PostgreSQL, Amazon RDS, ElastiCache: Redis / Valkey
- Cloud & Infrastructure: AWS (EMR, Lambda, SQS, Aurora), Docker, K8S, Datadog

## Experience

### Data Platform Lead, Lakehouse & AI (Jan 2025 – Present)

**Customer:** US adtech company

Lead the data platform and applied-AI initiatives for a US adtech company, owning lakehouse modernization, production reliability, and agentic AI systems while leading data engineering delivery.

Responsibilities:

- Spearheaded the end-to-end migration of a business-critical analytics platform from a legacy cloud warehouse and third-party analytical database to a Databricks Lakehouse, establishing a unified serving layer and removing vendor dependencies from the critical data path.
- Architected medallion-style data pipelines using Spark and Delta Lake: incremental ingestion, snapshot-consistent merges, granular fact models, pre-aggregations, and probabilistic counting for large-scale analytical workloads.
- Led controlled cutovers across data pipelines, backend APIs, frontend clients, and downstream services using automated source-versus-target parity testing and stable interface contracts; retired obsolete infrastructure and removed thousands of lines of legacy code.
- Delivered substantial efficiency improvements: 4–5× faster production processing, 100×+ acceleration in validated analytical prototypes, order-of-magnitude corrections to sizing logic, and recurring cloud-cost reductions.
- Restored correctness in complex measurement and attribution pipelines by resolving temporal-window, slowly-changing-dimension, canonicalization, deduplication, status-management, and aggregation defects.
- Led root-cause analysis and remediation of high-impact production incidents spanning schema inference, Unicode normalization, database connection lifecycle, distributed storage configuration, cache semantics, and accidental Cartesian joins.
- Built and launched an LLM-powered natural-language data discovery service from zero to production using FastAPI, agent frameworks, MCP tools, embeddings, and vector search over tens of thousands of domain attributes.
- Designed and productionized a Claude-powered autonomous incident-triage agent that performs permission-gated, read-only investigations, maintains an auditable activity trail, posts structured incident summaries, and opens draft documentation pull requests.
- Led the architecture (not implementation) of a resilient, self-learning agent-based LLM parser for adtech postlogs: scoped the stack, intermediate-result storage, off-the-shelf vs hosted OCR / bounding-box models (e.g., Datalab), clear responsibility separation from other agents in the flow, and an assessment of current SOTA approaches.
- Hardened agentic systems for production with structured-output validation, permission boundaries, bounded concurrency, token controls, observability, secrets management, container orchestration, and Git-based episodic memory.
- Applied Claude and Claude Code as engineering force multipliers for repository-scale analysis, migration planning, debugging, implementation, and documentation, backed by automated tests, data-parity checks, and human review.
- Automated several data-team and sprint/planning processes as reusable agent skills, converting runbooks into codified workflows with built-in confirmation steps and invariant checks.

### Messaging Project (Oct 2021 – Dec 2024)

**Customer:** US SaaS company

Developed core components of omnichannel messaging platform, enabling enterprise customers to reach users across WhatsApp Business, RCS, and SMS channels through unified APIs.

Responsibilities:

- Implemented key parts of sender registration system supporting multiple messaging channels (WhatsApp, RCS and potentially others), handling provider-specific requirements and compliance rules.
- Optimized message delivery latency and reliability while maintaining complex business rules across different OTT providers.
- Resolved critical customer escalations through deep technical investigation and edge case analysis.
- Worked on updating billing logic on a critical path, ensuring accurate and timely billing for customers.
- Contributed to cross-team Scala, Golang and Java projects.

### Marketing Project (Mar 2021 – Oct 2021)

**Customer:** US technology company

Developed and optimized Apache Spark pipelines processing cross-service engagement data with strict privacy preservation requirements across multiple digital entertainment and subscription platforms.

Responsibilities:

- Engineered high-performance Spark jobs processing TB-scale user engagement data.
- Built privacy-preserving data aggregation pipelines enabling anonymous cross-service analytics.
- Optimized data processing pipelines reducing job completion times while ensuring data minimization principles.
- Documented dataset lineage and data flow for compliance and reproducibility.

### Healthcare Project (May 2020 – Mar 2021)

**Customer:** US healthcare technology company

Played key role in modernizing healthcare data processing platform, enabling efficient transformation of diverse medical records into analytics-ready formats for business intelligence.

Responsibilities:

- Transformed legacy Python batch jobs into scalable Apache Spark workflows, migrating complex healthcare data processing logic.
- Designed and implemented production-grade ETL pipelines handling diverse healthcare data formats from multiple source systems.
- Optimized large-scale data reprocessing jobs reducing execution time while ensuring HIPAA compliance.
- Developed automated AWS S3 to Redshift data pipeline using PySpark and boto3, enabling real-time BI reporting.
- Contributed significant technical input to architecture decisions affecting multiple project initiatives.

### Adtech Project (Oct 2019 – May 2020)

**Customer:** Israel adtech company

Developed real-time ad analytics platform processing high-volume impression data (670+ MPS) to enable automated bidding decisions and campaign optimization.

Responsibilities:

- Implemented Spark Structured Streaming jobs for real-time ad performance analysis
- Optimized data processing architecture reducing operational costs while maintaining system reliability
- Modernized legacy Python ETL pipelines to improve maintainability and processing efficiency

### Communications Project (Feb 2018 – Oct 2019)

**Customer:** US SaaS company

Developed high-throughput streaming applications for SMS data processing, implementing complex rate calculation logic and generating business intelligence insights.

Responsibilities:

- Optimized messaging metadata enrichment pipeline during Kinesis-to-Kafka migration, achieving 830+ MPS through elimination of processing redundancies.
- Implemented comprehensive cross-stream validation to ensure data consistency between source systems during the transition.
- Built cost-effective internal services replacing external API dependencies, resulting in significant operational savings
- Led Scala knowledge-sharing initiatives including mentoring sessions and technical workshops

### Digital Transformation Project (Jun 2017 – Dec 2017)

**Customer:** US energy company

Contributed to enterprise-wide data modernization initiative, migrating traditional database workloads to cloud-based big data processing platform.

Responsibilities:

- Replaced legacy Oracle and MySQL batch processes with scalable Apache Spark pipelines
- Built new data processing workflows using Spark SQL and Hive, eliminating dependency on legacy SQL jobs

### IoT Monitoring Platform (Nov 2015 – Jun 2017)

**Customers:** European food and beverage companies

Developed real-time monitoring system processing data from IoT sensors across warehouse and retail locations, enabling predictive maintenance and inventory optimization.

Responsibilities:

- Implemented scalable data processing pipelines using Apache Spark and Akka Streams to handle real-time sensor data
- Built diagnostic tools enabling QA and hardware teams to validate device performance and data accuracy
- Developed anomaly detection system for early identification of hardware issues

## Education

**MSc Degree, Computer Science**, Lviv Polytechnic National University, Ukraine (2016 – 2017)

## Other Professional Development

### Natural Language Processing & AI (2023)

- Contributed Ukrainian localization and dataset improvements to OpenAssistant/oasst1 open-source project
- Built practical applications using transformer models and HuggingFace libraries
- Explored large language models and their enterprise applications

### Deep Learning & Time Series Analysis (2022)

- Completed fast.ai's Practical Deep Learning for Coders course
- Implemented time-series forecasting solutions using Facebook's Prophet
- Applied deep learning techniques to real-world data problems

## Personal Projects

I enjoy designing practical, custom-built apps with Claude and OpenAI Codex:

- A real-time FTMS logging and workout tracker for my air bike, with heart-rate and cadence support and data export.
- A high-performance 180° and 360° panorama and stereo-image viewer for the Meta Quest VR headset, with mipmap support and low latency.
- Automation of personal workflows with a Nous Hermes agent, through custom-built skills and purpose-built mini-apps.

## Languages

**English** - Advanced