Location: Latvia or Lithuania (Hybrid) • Employment: Full‑time • Reports to: Co‑founder & CTO

About Evergrowth

Evergrowth is a well‑funded startup with strong traction, backed by the co‑founder and former CEO of Trustpilot and a set of US + EU investors. We build AI systems that drive measurable business outcomes for our customers. We're growing and we're hiring a senior AI engineer to own agent‑style workflows end‑to‑end: from problem framing and prompt design to Python implementation, shipping to production, and continuous optimization.

The role

You'll be the primary builder of our Python backend that orchestrates "agents" (tool‑calling workflows powered by LLMs). "Agent" here means: APIs + smart prompting + context management + robust workflows that actually deliver business results. You'll work directly with the CTO on architecture, workflows, and tooling, and you'll have wide latitude to choose the right approaches for cost, performance, and maintainability.

What you'll do

• Translate business goals into concrete AI Agentic workflow, prompts and predictable outputs
• Design, implement, and ship agent workflows in Python
• Write excellent prompts and structured output schemas (JSON/JSON‑Schema, Pydantic) that are robust across models and providers.
• Integrate and operate multiple LLM providers (OpenAI primarily, also Anthropic and Google today, evaluate Groq/Cerebras and others for cost/latency/perf).
• Build observability for LLM calls: tracing, logging, cost accounting, and experiment tracking
• Make architecture choices with the CTO: orchestration patterns, state machines vs. DAGs, caching/memory, vector stores, background jobs.
• Continuously improve reliability and safety: guardrails, validation, rate limiting, content filters, and PII handling.
• Collaborate with product/ops to validate outputs with users, and iterate quickly.

Our current/target stack (you don't need all of this)

• Python: 3.12+, FastAPI, Pydantic, asyncio/httpx, pytest, mypy
• LLM orchestration & evals: OpenAI/Anthropic/Gemini; Orq.ai and LangSmith are in use today (nice‑to‑have, not required); experience with lightweight, framework‑agnostic patterns is valued.
• Data & infra: SQL, Redis, managed vector DB, S3‑compatible storage, Docket, Sentry.
• Patterns: function/tool calling, retrieval‑augmented generation (RAG), structured outputs, self‑checks, retries/backoffs, circuit breakers, streaming, KV/cache management.
• Performance: batching, token/latency/cost monitoring, provider fallbacks, model selection/routing.

What 'great' looks like

• You think in systems: clear interfaces, testable components, and metrics‑driven decisions.
• You're fluent in Python and comfortable owning production services (code quality, tests, CI, monitoring).
• You can read a domain, craft prompts and tool specs that reflect it, and iterate based on evidence and sometimes vibes. :)
• You're pragmatic about frameworks: happy to use them when they help, and to go framework‑light when it reduces complexity.
• You keep an eye on the frontier (models, inference, tooling) and can separate durable patterns from weekly hype.

Requirements

• 4+ years of Software engineering with strong Python (shipping production systems).
• 1–2+ years building with LLMs in production (multi‑model/provider experience, not just demos).
• Hands‑on experience with tool/function calling, structured outputs, context management.
• Comfortable designing APIs, writing tests, instrumenting systems, and debugging failures across model/provider boundaries.
• Working proficiency in English; based in Latvia or Lithuania and able to work hybrid (home + office) on a flexible schedule.

Nice to have

• Experience with LangChain/LangSmith, Orq.ai, LlamaIndex, DSPy, RAG, Guardrails/Outlines/Instructor.
• Vector DBs, embeddings, and data pipelines.

Why join Evergrowth

• Direct impact and ownership: report to the CTO, shape architecture and tooling from day one.
• The mandate to build what works: cost‑aware, maintainable, and fast.
• Competitive salary, stake in the company, flexible hybrid schedule, and the support of top‑tier EU/US backers.

First 90 days - example outcomes

• Review and improve existing core Agentic workflows
• Improve the quality of existing LLM outputs through prompting
• Introduce auto Value Proposition feature to production
• Review and reduce costs of existing pipelines by at-least 50%

Hiring process (fast and respectful)

• Intro with CTO (role, your background, mutual fit).
• Technical deep‑dive (systems + Python + LLM design), discussion‑based.
• Practical working session (pair on a small agent/workflow; no unpaid marathon).
• References and offer.