Requirements:
- 4+ years of production Python (not notebooks): FastAPI/async, PostgreSQL, queues (Celery / RabbitMQ / SQS), Docker
- 2+ years with LLMs/VLMs in production, not side projects
- Hands-on with open-weight VLMs: Qwen2.5-VL / Qwen3-VL, InternVL, Llama Vision, Pixtral — understands their actual strengths and failure modes rather than having just read about them
- Structured extraction: JSON-schema output, constrained decoding, document workloads (scans, PDFs, forms), post-processing and validation
- Fine-tuning: LoRA / QLoRA, PEFT
- Deployment and serving: vLLM or SGLang / Triton, quantization (AWQ / GPTQ / FP8), batching, KV cache, sizing GPUs against an SLA and a budget
- Classical CV: OpenCV, layout detection (DocLayout, Surya), PaddleOCR / Tesseract, YOLO
- Agentic frameworks (LangGraph), RAG, vector databases
- MLOps: MLflow / W&B, CI/CD for models
- Presales or estimation experience
Soft skills:
- English level: Upper-Intermediate (B2) or higher
- Self-directed
- Comfortable working from ambiguous requirements
What you will get:
Career growth & projects
1. Competitive salary based on your skills and experience, and a benefits package.
2. Regular performance appraisals to support your career development.
3. Professional development support: online courses, certifications, and study compensation (including English courses).
4. A modern tech stack across challenging, long-term projects.
Health & well-being
Paid vacation (18 working days), public holidays (11 days), and sick leave (15 days).
VIP medical insurance or sports coverage.
Work environment & flexibility
Remote work from anywhere in the world or in one of our offices in Kyiv, Lviv, or Warsaw.
Necessary equipment to perform your work.
Culture & community
A collaborative and open culture where contributions are recognized.
Regular corporate online & offline activities.
Geniusee Charity Fund, making a difference beyond tech.
