About Arango
At Arango we're on a mission to make working with complex data simple, powerful, and AI-ready. Based in California and in Cologne (with a global team), we’re building a cutting-edge data platform that helps organizations bring all their data together — graph, document, key/value, full-text, and vector search — in one engine.
Why does that matter? Because it means developers and data teams can build next-gen AI applications like RAG, knowledge graphs, and smart agents — without gluing together a bunch of tools that were never meant to work together.
Our platform makes it easy to work with any kind of data-structured, semi-structured, or unstructured, and it gives teams everything they need to build faster, smarter, and with way more context. From our easy-to-learn AQL query language to modern integration tools, we’re here to help teams grow and scale with AI.
About The Role
Arango is looking for an AI Forward Deploy Engineer to embed with our customers and deliver real, production‑grade AI solutions—fast. You’ll run discovery, design and prototype systems, ship secure and reliable services, and ensure adoption and measurable business impact. Think full‑stack AI + MLOps + product sense, delivered side‑by‑side with users.
This is a hands-on, customer-facing role for engineers who are as comfortable whiteboarding with executives as they are profiling latency in a retrieval pipeline.
What you’ll do
Customer discovery & scoping
- Partner with customer sponsors, SMEs, and operators to identify high‑value AI use cases.
- Define success metrics, SLAs/SLOs, data access needs, and a delivery plan (PoV → pilot → production).
Prototype → production
- Build end‑to‑end prototypes (data connectors, RAG pipelines, prompts/tools, UIs/APIs).
- Productionize into secure, observable services with CI/CD, infrastructure‑as‑code, and proper testing.
LLM application engineering
- Implement retrieval‑augmented generation (chunking, embeddings, ranking, caching) and tool/call orchestration.
- Evaluate and iterate prompts, models, and retrieval strategies using offline/online metrics and A/B tests.
- Where needed, fine‑tune or adapt models (LoRA/PEFT, preference optimization/DPO, distillation) and optimize inference (quantization, batching, vLLM/TGI/TensorRT‑LLM).
Data, reliability & cost
- Build robust data pipelines (ETL/ELT), vector indices, and metadata governance.
- Monitor quality, drift, hallucination/guardrail events, latency, and cost; set up alerting and dashboards.
Security, privacy & compliance
- Implement role‑based access, secrets management, audit logging, PII redaction, and content safety filters.
- Align solutions to customer requirements (e.g., SOC2/ISO 27001, GDPR/CCPA, HIPAA as applicable).
Enablement & change management
- Document architectures and playbooks; train customer engineers and end users.
- Capture product feedback and influence [Company]’s roadmap with field learnings.
Qualifications
Required
- 4+ years of software engineering experience building and operating production systems (or equivalent).
- Strong AI, Python skills; solid understanding of data structures, networking, concurrency, and systems design.
- Hands‑on experience with modern LLMs and tooling (e.g., OpenAI/Anthropic/Llama, Hugging Face, LangChain/LlamaIndex, function/tool calling).
- Retrieval and vector databases (FAISS, pgvector, Pinecone, Weaviate, or similar).
- Cloud & containers (AWS/GCP/Azure), Docker/Kubernetes, IaC (Terraform/CloudFormation), and CI/CD.
- Observability (metrics, logs, traces) and performance tuning for latency‑sensitive services.
- Excellent communication; experience working directly with customers or cross‑functional stakeholders.
Nice to have
- Front‑end or full‑stack experience (TypeScript/React, Next.js) for light UI prototyping.
- Search/IR fundamentals (BM25, hybrid retrieval, re‑ranking).
- MLOps platforms (MLflow, Weights & Biases), evaluation frameworks (Ragas, promptfoo, DeepEval).
- Inference optimization (vLLM, Text Generation Inference, Triton, TensorRT‑LLM, quantization).
- Domain experience in [finance/healthcare/public sector/manufacturing/retail].
- Security/compliance familiarity; prior work with data residency, KMS/HSM, or private networking.
- Government/industry clearances where relevant.
What success looks like (6–12 months)
- 2–4 customer use cases deployed to production with agreed‑upon uptime, latency, and cost targets.
- Demonstrable quality lifts (e.g., task accuracy, deflection rate, cycle time) backed by evals and telemetry.
- Reusable building blocks (templates/operators/connectors) adopted by the broader delivery team.
- Customer enablement completed (runbooks, docs, training) with strong satisfaction/NPS.
Our toolset (you don’t need all of these)
- Models & SDKs: OpenAI, Anthropic, Meta Llama, Hugging Face
- Retrieval: FAISS, pgvector, Pinecone, Weaviate; rerankers (ColBERT, cross‑encoders)
- Pipelines & Orchestration: LangChain, LlamaIndex, Ray, Airflow
- MLOps & Evals: MLflow, Weights & Biases, Ragas, promptfoo, Great Expectations
- Serving & Infra: vLLM, TGI, FastAPI/gRPC, Docker/K8s, Terraform, GitHub Actions
- Observability & Guardrails: OpenTelemetry, Prometheus/Grafana, Llama Guard/Content Safety, custom filters
- Data: Postgres/BigQuery/Snowflake; Kafka; object storage
Sample projects you might tackle
- Secure RAG assistant over millions of documents with hybrid search, guardrails, and cost‑aware caching.
- High‑volume customer support copilot that integrates with CRM/ITSM and meets a strict response‑time SLO.
- Document intake pipeline with PII detection/redaction and structured extraction (LLM + regex + heuristics).
- Fine‑tuned classification/routing model that reduces human triage by 40%+
Powered by JazzHR
0NoGtmxGKV
