Who I Am
Artyom Kashuta. Full-stack and AI engineer. 7+ years running production systems — APIs, data pipelines, LLM/agent workflows, and cloud infrastructure. Remote-first.
Currently: AI Automation Engineer at TaxDome — designing and delivering end-to-end LLM/agent workflows in Python and TypeScript, building eval systems and retrieval pipelines over internal documents and CRM data.
Previously: Staff Engineer and Team Lead at 9RED (ex. Pari) — tech-led cross-domain platform architecture and numerous API integrations serving >15,000 MAU and 700 RPS per instance.
What I Do
- AI agent engineering. End-to-end workflows with LangGraph, LangChain, and LangSmith. Direct OpenAI / Anthropic / Gemini API calls where fine-grained token, latency, and cache control matters. Eval systems and prompt-regression suites via LangSmith traces, structured outputs, and golden datasets — to keep agent quality from degrading as prompts and models evolve.
- Retrieval & RAG. Chunking, embeddings, hybrid search, reranking over internal documents and CRM data. FastAPI services, consumer-side agents.
- Backend and distributed systems. Python (FastAPI), TypeScript / Node.js (Fastify), event sourcing on Apache Kafka, inter-service communication via gRPC, observability through OpenTelemetry → Prometheus → Grafana → Mimir.
- AI MVP development. Backend + lightweight frontend + AWS infrastructure in Docker — to quickly validate hypotheses and cut time-to-value.
What I’ve Measurably Built
- Platforms at 700 RPS with <150 ms latency for >15k MAU — Wallet and sportsbook/casino integrations at 9RED.
- Migrated 20+ microservices from Kafka / GraphQL Federation to gRPC contracts → −32% inter-service latency.
- Full observability from scratch (OpenTelemetry → Collector → Prometheus → Grafana, plus Mimir for traces) → −45% MTTR.
- Event sourcing on Apache Kafka across 20+ services → +50% scalability.
- Led cross-functional team of 8 engineers (3 BE, 2 FE, 3 QA) and 1 BA.
Stack
- Languages: Python, TypeScript / Node.js, JavaScript, SQL (PostgreSQL / PostGIS).
- Backend: FastAPI, Fastify, Express; microservices, event sourcing, gRPC, REST, GraphQL, webhooks.
- AI / LLM: OpenAI, Anthropic, Gemini APIs; LangGraph, LangChain, LangSmith (evals & tracing); RAG, embeddings, vector & hybrid search; agentic frameworks; prompt engineering. Claude Code and Cursor — daily drivers. n8n, Make.
- Data: PostgreSQL, PostGIS, MongoDB, DynamoDB, Redis; Apache Kafka, RabbitMQ; ORMs — Prisma, Sequelize, Drizzle.
- Cloud / observability: AWS (EKS, S3, RDS, CloudWatch), Docker, Kubernetes, GitLab CI/CD, GitHub Actions; OpenTelemetry, Prometheus, Grafana, Mimir.
Detailed list with rationales — on the /uses page.
What I’ve Written
- Claude Code Guide — 14-lesson course on Claude Code internals: harness, context window, skills, hooks, MCP, subagents, models, antipatterns. RU + EN, interactive exercises, certificate on 100% completion.
- artka.dev — this site. Astro 5 + PostgreSQL + Drizzle, bilingual RU/EN, SSG + dynamic admin panel, Mermaid and LaTeX rendering at build-time, on-demand PNG certificates via Satori.
- AI agent engineering writeups — postmortems and breakdowns of agentic systems: tool design, evaluation, production failure modes.
What I’m working on right now — on the /now page.
Education and Languages
- Elbrus Coding Bootcamp — Full-stack Developer Certificate, Moscow, 2019–2020.
- English — Advanced (C1) · Russian — Native.
Contact
- Email: a@artka.dev
- GitHub: github.com/node-develop
- LinkedIn: linkedin.com/in/artem-kashuta
- X: @artkadev
- Telegram: @akv6020
- RSS: RU · EN