AI/ML Platform Engineer
Lekta AI
This role is about owning the entire AI/ML platform that powers Lekta's conversational agents. You will be responsible for integrating and optimizing language models (LLMs), speech models (ASR/TTS), retrieval-augmented generation (RAG), and evaluation frameworks. The work is driven by the need for ultra-low latency in voice channels and reliability in regulated domains (banking, insurance, telco). You will make architectural decisions between deterministic logic and neural components, and stay on top of the AI frontier to guide adoption of new techniques. It is a hands-on engineering role, not research: you ship production code in TypeScript or Python and use AI tooling (Claude Code) daily.
Brakuje: brak informacji o wielkości zespołu platformowego, nie określono, czy jest dyżur on-call.
This role is about owning the entire AI/ML platform that powers Lekta's conversational agents. You will be responsible for integrating and optimizing language models (LLMs), speech models (ASR/TTS), retrieval-augmented generation (RAG), and evaluation frameworks. The work is driven by the need for ultra-low latency in voice channels and reliability in regulated domains (banking, insurance, telco). You will make architectural decisions between deterministic logic and neural components, and stay on top of the AI frontier to guide adoption of new techniques. It is a hands-on engineering role, not research: you ship production code in TypeScript or Python and use AI tooling (Claude Code) daily.
- ✓AI-first workflow with company-paid tooling (Claude Code, model APIs)
- ✓Direct impact on real customer experience in production
- ✓Low bureaucracy, results-oriented culture
- ✓Honest conversation about trajectory as company scales
- !No mention of team size or on-call responsibilities
- !Small company (26-50) could mean less process and more chaos
- •Integrating and evaluating frontier language models (provider APIs, prompting, cost-quality trade-offs)
- •Tuning ASR and TTS pipelines for latency, multilingual quality, and production reliability
- •Implementing incremental dialogue management and end-of-turn detection for real-time voice
- •Building and maintaining RAG pipelines (embeddings, hybrid retrieval, recall-precision optimization)
- •Developing and extending evaluation frameworks to catch regressions and monitor drift
- •Making architectural decisions on what belongs in deterministic logic vs. language model invocations
- •Reviewing and shipping production code, using AI assistants (Claude Code) to accelerate delivery
- •Tracking model releases and new techniques, deciding what to adopt for the platform
Oferta dla doświadczonych specjalistów (Senior).
A mid-level engineer with at least 2-3 years of production experience in AI/ML, comfortable with LLM integration and basic speech pipelines, but willing to learn the rest. They should have strong software engineering fundamentals and be able to ship production code, even if they haven't yet specialized in all areas listed.
Junior developers without production AI experience; engineers who prefer research over shipping; those uncomfortable with the pace and ownership of a small company (26-50); people who cannot work in Polish (C1 required).
- ?Ile osób liczy zespół inżynieryjny?
- ?Czy jest dyżur on-call? Jak często?
- ?Jak wygląda obecny stos technologiczny platformy (np. jakie modele, jakie narzędzia do ewaluacji)?
- ?Czy istnieje już platforma, czy budujemy od zera?
- ?Jak mierzycie sukces na poziomie platformy?
- ?Jaki jest priorytet: redukcja kosztów, poprawa jakości, czy szybkość wdrażania?
- ?Czy oferujecie budżet na konferencje lub szkolenia?
- ?Jak wygląda proces decyzyjny przy wyborze nowych modeli lub technik?
- −Brak informacji o wielkości zespołu platformowego
- −Nie określono, czy jest dyżur on-call
- −Brak szczegółów dotyczących obecnego stanu platformy (np. jakie modele już są używane)
- −Nie podano narzędzi do ewaluacji (ramowe, własne?)
- −Brak informacji o budżecie na rozwój (szkolenia, konferencje)
Wynikowa kultura jest zorientowana na wyniki, niska biurokracja, z silnym naciskiem na efektywność i autonomię. Zespół używa AI jako narzędzia w codziennej pracy, co sprzyja szybkiemu prototypowaniu i wdrażaniu.