Applied AI Engineer (Document Intelligence and Prompting) в потужний стартап to $7000

Вітаю тебе, фахівець \0/

Мене звати Кирил, я допомагаю зібрати дрім тім для своїх друзів з Hooh – це стартап від фаундерів Depositphotos. Команда сіньорна та досвідчена, тому опортʼюніті, ІМХО, дуже кльове.

отже

Hooh – це b2c AI-продукт, який робить складні документи простими для розуміння та дій. 

Він дозволяє завантажувати файли (PDF, фото, скани), автоматично розпізнавати їхній вміст, витягувати ключову інформацію та отримувати короткі, зрозумілі підсумки.
 

Ми успішно випустили v1, підтвердивши нашу основну AI-пайплайн-архітектуру. Зараз ми розширюємо продукт до “scanner-grade” PDF-досвіду, що поєднує глибоке AI-розуміння з надійним редагуванням документів.

Далі дозволю собі перемкнутись на англійську, бо мова піде про конкретні речі.
 

_________________

CEO about the role: “We are looking for an Applied AI Engineer who will help us make messy documents reliably understandable, build accurate structure and entity extraction, strong retrieval, and cost‑efficient model orchestration."

Key Responsibilities

• Document analysis & taxonomy: Systematically analyze layouts/sections/entities/relations; define categories, types, and metadata schemas that generalize across domains.
• Prompting & schemas: Author, version, and maintain prompts/chains for classification, extraction, summarization, and Q&A; enforce JSON/JSON‑Schema outputs.
• Retrieval (RAG): Build hybrid retrieval (embeddings + keyword + graph traversal); manage embedding generation, indexing, dedupe, and drift.
• Cascades & MoE: Architect model cascades (fast filters → experts → fallback LLMs) and MoE routing to balance quality/cost/latency.
• Agentic orchestration: Use CrewAI (or similar) to coordinate multi‑step pipelines; add caching/batching/streaming and robust retries.
• Result fusion: Combine outputs via reranking, voting, and confidence/consistency checks; add guardrails and safety rules.
• Evaluation & monitoring: Run LangSmith/Evals (or similar), define golden sets, automate regressions, track quality/cost/latency; canary and observe.
• Operate models locally/hosted: Work with vLLM/TGI/Ollama/llama.cpp as well as API models; apply quantization/LoRA when useful.
• Data stores: Use vector DBs plus graph and relational DBs to power retrieval and joins (e.g., Qdrant/Weaviate/Milvus/pgvector + Neo4j + PostgreSQL).

Required Skills & Experience

• Strong programming skills in Python; ability to write clear, tested, maintainable code.
• Experience building at least one LLM‑powered feature end‑to‑end (prototype → production), or equivalent open‑source/portfolio work.
• Hands‑on with RAG, vector embeddings, and evaluation (offline + online A/B, error analysis).
• Familiarity with model cascading/MoE concepts (or ability to learn quickly).
• Practical database skills: comfortable with SQL and with either graph or vector systems.
• Product mindset: bias to measure impact and iterate.

We hire for capability and learning speed. If you’re strong on fundamentals and can show work, we’d love to talk, even if you don’t check every box.
 

Nice to have:

• OCR/Document AI, ontology design, labeling workflows.
• LangSmith/Evals, LangChain/LlamaIndex; Ray/Airflow; GPU basics.
• Security & privacy for AI systems (PII handling, GDPR), prompt/response guardrails.

The Team

• CEO
• Product Manager
• Fullstack Engineering Lead
• Back End Dev
• Fullstack Dev 
• AI engineer
• User Acquisition Manager
• 2х Designers 
• 2x iOS Devs
  
   

What We Offer

• Challenging technical problems and the environment to solve them.
• Flexible schedule and fully remote work from anywhere with reliable overlap to CET.
• 18 paid vacation days, 8 no-paperwork sick days, and 11 public holidays per year.
• Minimal bureaucracy, no micromanagement, and swift decision-making.
• Direct influence on strategy and the ability to see your experiments ship fast.
• Be part of a team that previously built and scaled successful businesses.

Hiring Process:

1. Intro Call
2. Interview with CEO
3. Offer