30.09.2026 aktualisiert
100 % verfügbar
Senior Software Engineer - AI & LLM Systems
Riga, Lettland
Nur Remote
Über mich
I build AI and LLM systems that hold up in production: RAG, semantic search on pgvector, agents with tool use, and the full stack around them in TypeScript, Node and Next.js. 20+ years shipping software, from a national vaccination portal with 1M+ users to multi-tenant SaaS. Riga-based, remote.
Skills
LangChainSupabaseJavaScriptWeb ScrapingPythonPostgreSQLNode.jsNext.jsTypeScriptReact.jsLarge Language ModelsPrompt EngineeringGoogle FirebasePuppeteer (Software)React Native
AI & LLM Systems
Expertise in RAG, semantic search, embeddings, and vector search in PostgreSQL (pgvector). Development of similarity thresholds, drift control, and deduplication for production-grade AI systems.
Agents & Tool Use
Design and implementation of LLM tool calling, single-tool and multi-agent pipelines with role definition and handoff, server-side verification of model output, and conversation state management.
Back-end & Data Engineering
Proficiency in Node.js, Express, tRPC, PostgreSQL, pgvector, Prisma, Drizzle, Supabase, Firebase, REST APIs, and OCPP for building scalable back-end services and data pipelines.
Structured Output & Validation
JSON schema enforcement, Zod and Typia validation, and downstream-safe parsing to ensure reliable and type-safe data flows.
Orchestration & Automation
Use of LangChain, custom Node/TypeScript agent pipelines, and n8n (self-hosted, 50+ node workflows) for complex automation and orchestration tasks.
Front-end Development
Building modern UIs with React, Next.js (App Router), React Native, Tailwind CSS, shadcn/ui, Radix UI, TanStack Query, and ECharts.
Infrastructure & DevOps
Deployment and management using Docker, CI/CD with GitHub Actions, Linux/self-hosted VPS, Nginx, Vercel, and Cloudflare.
Evals & Observability
Creation of benchmark test sets, output quality and cost tracking, and Langfuse tracing for monitoring and improving AI system performance.
Automation & Platform
Large-scale data ingestion, Puppeteer, Playwright, Turbo, and multi-tenant SaaS platform development.
Programming Languages
Strong command of TypeScript, JavaScript, Python, SQL, and HTML/CSS for full-stack and AI-focused development.
Expertise in RAG, semantic search, embeddings, and vector search in PostgreSQL (pgvector). Development of similarity thresholds, drift control, and deduplication for production-grade AI systems.
Agents & Tool Use
Design and implementation of LLM tool calling, single-tool and multi-agent pipelines with role definition and handoff, server-side verification of model output, and conversation state management.
Back-end & Data Engineering
Proficiency in Node.js, Express, tRPC, PostgreSQL, pgvector, Prisma, Drizzle, Supabase, Firebase, REST APIs, and OCPP for building scalable back-end services and data pipelines.
Structured Output & Validation
JSON schema enforcement, Zod and Typia validation, and downstream-safe parsing to ensure reliable and type-safe data flows.
Orchestration & Automation
Use of LangChain, custom Node/TypeScript agent pipelines, and n8n (self-hosted, 50+ node workflows) for complex automation and orchestration tasks.
Front-end Development
Building modern UIs with React, Next.js (App Router), React Native, Tailwind CSS, shadcn/ui, Radix UI, TanStack Query, and ECharts.
Infrastructure & DevOps
Deployment and management using Docker, CI/CD with GitHub Actions, Linux/self-hosted VPS, Nginx, Vercel, and Cloudflare.
Evals & Observability
Creation of benchmark test sets, output quality and cost tracking, and Langfuse tracing for monitoring and improving AI system performance.
Automation & Platform
Large-scale data ingestion, Puppeteer, Playwright, Turbo, and multi-tenant SaaS platform development.
Programming Languages
Strong command of TypeScript, JavaScript, Python, SQL, and HTML/CSS for full-stack and AI-focused development.
Erfahrung
Berufserfahrungmehr als 10 Jahre
Freelancer seit2005
Sprachen
EnglishverhandlungssicherLatvianMuttersprache
Projekthistorie
Turned recorded support calls into structured, searchable records backed by a self-growing catalog of known issues. Implemented semantic matching for real spoken language, server-side scoring of new phrasings, and a live product that predicts a caller's likely issue before anyone picks up. Stack: OpenAI embeddings, PostgreSQL + pgvector, RAG, LLM tool use, Chirp STT, Node.js, TypeScript.
Built the AI engine that turns a single property listing into 60+ social posts across 6 platforms over a 6-week campaign. Re-architected generation from n8n workflow to a native in-API service, automated the full path from generation to approval to scheduled posting, and rebuilt the agent dashboard in Next.js 15. Stack: Next.js 15, React, TypeScript, Supabase, PostgreSQL, LLMs, prompt engineering, MLS Grid API, Node.js.
Shipped a multi-tenant CRM/ERP covering intake, full consultation workflow from booking to sign-off, clinical notes, multi-currency invoicing, and growth charts. Type-safe from database to interface with full client separation and audit trail. Stack: Next.js 14, TypeScript, tRPC v11, Drizzle ORM, PostgreSQL, Clerk, Zod, Typia, Tailwind, shadcn/ui, ECharts.