The posting
📍 Location: Remote (PL / EU)
💰 Compensation (B2B): To be agreed (proposed 25–35k PLN net B2B). All travel expenses covered.
📝 Contract Type:Full-time • B2B
📈 Seniority Level: Senior / Principal
🧰 Tech Stack Tags:
Python · FastAPI · Docker · K8s · LangChain · LlamaIndex · LLM · HuggingFace · Cursor · Windsurf · Qdrant · gRPC · Azure/AWS · MLOps
🏢 About Us
Join a team building some of the most advanced AI systems in the world.
We develop core IT solutions supporting federal agencies and law enforcement in the U.S., where technology truly makes a difference.
Our systems leverage:
- the latest advancements in natural language processing (NLP),
- multimodal AI analyzing text, images, and sound,
- and intelligent decision automation enabling real-time response.
These are systems that detect threats faster, analyze evidence more efficiently, and support humans in making the right calls—when every second counts.
Our solutions run in high-security environments:
air-gapped or in private cloud deployments, compliant with CJIS and FISMA standards—ensuring full data privacy and integrity.
🎯 What You’ll Do
- Design and maintain microservices (REST/gRPC) exposing AI models via API
- Package and deploy models (vLLM, Ollama, HF Transformers) with Docker + CI/CD
- Build advanced retrieval pipelines combining RAG and hybrid dense‑sparse search (BM25 + embeddings, multi‑vector, re‑ranking) using Qdrant/Weaviate/Pinecone/Elasticsearch and frameworks like LangChain, LlamaIndex, LangGraph.
- Perform fine-tuning, quantization, and RLHF/RLAS on Llama 3/4, Qwen, Mistral (datasets, experiments, evaluation)
- Create prompt tooling (templates, guardrails, eval harnesses) and benchmarking aligned with CJIS requirements
- Implement model quality monitoring (latency, token cost, hallucination rate) and security policies (Guardrails AI, LlamaFirewall)
- Integrate with SIEM/SOC systems and databases (Elasticsearch, PostgreSQL)
- Co-develop MLOps standards, testing, logging, observability (Prometheus, Grafana)
- Work with security teams on audits and certifications (CJIS, FedRAMP)
- Optional business travel to Japan (Roppongi Hills, Tokyo) and the United States (Irvine, California).
🔧 Must-Haves
- 8+ years Python 3.x (FastAPI / Django / Flask)
- 2+ years hands-on with LLMs (OpenAI, Azure OpenAI, Anthropic) and self-hosted models (Llama 3/4, Mistral)
- Experience building RAG, hybrid search systems and agents using LangChain / LlamaIndex / LangGraph
- Experience with Vector DBs (Qdrant / Milvus / Pinecone) and agent frameworks (LangGraph, CrewAI)
- Docker (multi-stage) + Kubernetes, including Helm or Kustomize
- Strong understanding of HTTP/REST, CI/CD (GitHub Actions / GitLab CI), and security-by-design principles
🌟 Nice-to-Haves
- Helm / Kustomize, Terraform
- Azure (AKS, OpenAI Service) or AWS (SageMaker, Bedrock)
- gRPC, event-driven architecture (Kafka / RabbitMQ)
- Experience with Guardrails AI, LlamaFirewall, PyTorch Lightning
🤝 Soft Skills
- Technical ownership and attention to documentation
- Clear communication in a distributed team (ENG B2+)
- Focus on security, performance, and regulatory compliance
🎁 What We Offer
- Real impact — your code will support law enforcement in fighting crime and saving lives
- 100% remote work + flexible hours (core overlap 10:00–15:00 CET)
- Access to a private LLM environment + GPU cluster (for feature-flagged demos)
- Training budget + participation in international Python/AI conferences



