full-stack AI engineer for internal platform

TrafficStars' internal product, IDP, is used daily by engineering, sales, performance, compliance, and other teams to run their work. It includes AI assistants, ad creative moderation, analytics tools, agent workstations, company knowledge search, and LLM call tracking.

• Ship features end to end, from schema and API through background jobs and UI to production
• Build AI agents and assistant tools, retrieval, streaming, and evaluation, while keeping token costs under control
• Test new models, agent SDKs, and MCP servers; measure results and decide whether to ship or discard them
• Connect AI to Postgres, ClickHouse, internal APIs, Slack, email, and LLM providers through OpenRouter
• Run long-running and scheduled work on Trigger.dev, including reports, checks, and syncs, with retries that avoid duplicate sends
• Review pull requests from several teams and their agents, catching permission gaps, data leaks, heavy queries, and migration clashes
• Improve coding agents with skills, rules, and CI checks
• Ship migrations and Kubernetes deployments, and roll back when needed

• At least 3 years of experience building real products, including full-stack production work
• Be curious about new models, agent frameworks, and tools, and distinguish useful developments from hype
• Communicate clearly, turn vague requests into working first versions, explain trade-offs to non-engineers, and provide constructive review comments
• Take ownership, clarify requirements, ship, listen, and iterate without waiting for a perfect specification
• Have shipped LLM features to real users, including tool calling, streaming, context limits, retries, caching, and cost control
• Apply practical AI safety principles, including protection against prompt injection, keeping personal data out of prompts and logs, and limiting tool permissions
• Evaluate AI quality using test sets and regression checks
• Use AI coding agents daily, such as Claude Code, Codex, or Cursor, and know when to delegate to them and when not to trust them
• Be strong in TypeScript, React, and Next.js, with the ability to design features from scratch
• Be comfortable with Postgres schemas, migrations, indexes, and query performance
• Understand Kubernetes and Docker well enough to deploy, inspect failing pod logs, and roll back
• Know security basics, including access control, injection risks, and secret handling
• Be proficient with Git, including resolving complex merges across parallel branches
• Будет плюсом: Supabase, ClickHouse or other analytics databases, OpenTelemetry, Grafana, debugging through traces, experience building internal tools or developer platforms, ad-tech experience with advertisers, publishers, or ad formats

• A whole company relies on shipped work from day one; there is no need to fight for traffic or a first customer
• Access to the latest AI models and agents
• Features can go from idea to production in days, with colleagues using them the same week
• Tools and agent improvements help dozens of people build faster
• Significant autonomy and influence over how AI is used across the company