AI Engineer

📋 Description • Join a fast-moving R&D powerhouse that turns private-equity vision into production-grade AI. As our AI Engineer you will own the end-to-end lifecycle of multiple AI products across ski-resort operations, medical clinics, and enterprise CRMs—impacting thousands of real users every day. • Architect and ship LLM-centric features that understand, summarize and reason over domain-specific documents (contracts, medical records, guest feedback) in English, Ukrainian and German. You will choose the right model (GPT-4, Claude, Llama-3, Mistral), design prompt chains, implement RAG with hybrid search (dense + sparse) and tune retrieval windows to hit <800 ms latency at 95 % accuracy. • Build autonomous agent networks that plan multi-step workflows—call APIs, query SQL/NoSQL, post updates to Slack and generate follow-up tasks—while exposing decision traces for human oversight. You will leverage LangGraph, LlamaIndex workflows or AutoGen to coordinate specialist agents and guarantee idempotent, recoverable execution. • Extend solutions beyond text: integrate vision transformers for on-slope hazard detection, OCR + layout models for form digitization, and speech-to-text for real-time doctor dictation. You will containerize each modality, expose gRPC/REST endpoints, and implement GPU/CPU autoscaling on AWS EKS/GKE to keep cloud cost per 1 000 inferences under budget. • Establish rigorous eval & observability: design unit tests, regression suites and LLM-as-a-judge benchmarks; capture latency, token cost, hallucination rate and user feedback in Grafana/Prometheus; trigger automatic rollback when KPIs drift. Your dashboards will be the single source of truth for executives and product managers. • Collaborate in a cross-functional pod—product lead, UX designer, DevOps, data engineer and security architect—in two-week sprints. You will present spikes, demos and decision records, turning research papers into Monday-morning prototypes and Friday-afternoon releases. • Champion ethical, secure AI: implement PII redaction, bias audits and EU AI-Act-compliant documentation. You will maintain model cards, data sheets and risk logs so every portfolio company can pass investor due-diligence and regulatory review. • Mentor junior engineers and run internal “AI Guild” sessions, sharing hacks on quantization, LoRA fine-tuning, and cost-effective GPU spot instances. Your upskilling culture will keep the bureau at the cutting edge and attract top talent across Europe. • Enjoy freedom to experiment: 20 % of your time is reserved for exploring neuromorphic chips, multi-modal embeddings or synthetic-data generation. Successful proofs-of-concept receive immediate funding and a green light for production pilots. 🎯 Requirements • 3+ years production experience designing, shipping and monitoring LLM-based systems (RAG, agents, fine-tuning) with measurable business impact. • Expert Python fluency and hands-on command of LangChain, LangGraph, LlamaIndex, Hugging Face Transformers, OpenAI/Anthropic SDKs; comfortable reading research papers and converting them into code within days. • Solid DevOps mindset: Docker, Kubernetes, CI/CD (GitHub Actions), infrastructure-as-code (Terraform/CDK); deployed GPU workloads on AWS, GCP or Azure and optimized for cost, latency and autoscaling. • Demonstrated multi-modal projects—vision (CLIP, DETR), speech (Whisper, wav2vec2)—including dataset creation, model fine-tuning and performance evaluation; can explain trade-offs between cloud APIs vs. self-hosted models. • Upper-Intermediate English and proactive communication: write clear design docs, present to investors, give/take feedback in a remote, async team across EU time-zones. 🏖️ Benefits • Fully remote, flexible schedule—work from any European location with a €2 000 home-office stipend and quarterly on-site hackathons in the Alps or Kyiv. • Unlimited GPU/cloud budget for experiments; personal learning wallet of €3 000/year for courses, conferences and academic journals. • Stock-option pool across the entire PE portfolio: when any of our startups exit, you share in the upside; historical IRR >35 %. • 25 days paid vacation plus Ukrainian public holidays, paid volunteer days, and mandatory “AI detox” long weekend every quarter to avoid burnout.