Applied AI Engineer

Location

Fully Remote (EU timezone overlap preferred)

Contract Type

Permanent

We are partnering with a tech scale-up to find a top-tier Applied AI Engineer . Our client is heavily investing in governed AI agents designed to work directly with customers and internal teams.

Data and Intelligence sit at the very center of their product ecosystem. Their platform is evolving to carry advanced intelligence: real-time decisioning, predictive modeling, and governed AI agents. They need an exceptional engineer to drive this shift.

If you want to build production agents that answer real business questions and act with real money in a high-stakes, highly governed environment, this is the role for you.

About the Role

Our client’s decision engine is governed by a formal decision register: over 100 decisions across 15 platform modules, spanning real-time risk gates, reward orchestration, payment routing, and responsible gambling interventions.

As an Applied AI Engineer on the Intelligence team, you are the product engineer of the data and agent platform. You will build production agents that are grounded, auditable, secured against adversarial input, and gated by human approval. Your first major deliverable will be a production SQL BI analyst agent: a Slack-native agent that answers executive business questions with governed SQL, validated queries, and cited evidence.

From there, the role balances four crafts in roughly equal measure:

1. AI Agents

• Analyst Agents:

Build and own Slack-native agents end-to-end that translate natural-language questions into governed SQL over the analytical warehouse.

• Customer-Facing Agents:

Develop triage and routing systems, retrieval-grounded (RAG) responses over versioned knowledge bases, and multi-turn conversational state machines with strict escalation logic integrated into helpdesk/CRM platforms.

• Risk-Stratified Tooling:

Build the layer between agents and back-office APIs. Harden agents against prompt injection, tool misuse, and data exfiltration. Design multi-turn confirmation workflows where low-risk actions graduate to full autonomy as evidence accumulates.

• Autonomy Frameworks:

Build using LangGraph, Anthropic Agent SDK, Model Context Protocol (MCP), or equivalent. Engineer closed feedback loops, decision audit logging, and strict evaluation harnesses.

2. Machine Learning

• Production Models:

Build and productize models for churn, LTV, bonus sensitivity, and composite player risk scores (identity, payment, gameplay, bot-play, multi-accounting).

• Governed Signals:

Ship models as versioned, SLA-backed contracts with the decision engine—served across real-time (Kafka/MSK) and batch (ClickHouse) tiers, complete with drift monitoring and automated retraining.

3. Data Science

• Risk Scoring:

Own multi-vector withdrawal risk scoring, including cited rationale, evidence-aware aggregation, and automatic re-scoring.

• Business Rules:

Codify business rules with domain owners. Simulate and backtest every threshold change against historical data, designing holdouts and control groups to measure true uplift.

4. Dashboards \& Surfaces

• Supervisor Surfaces:

Build review queues, one-click action proposal cards for high-risk mutations, and searchable session replays exposing prompts, reasoning chains, and tool calls.

• Business Dashboards:

Design KPI views and decision audit dashboards that move toward AI-assisted anomaly detection, built in collaboration with the BI team.

About You

• Experience:

4+ years in software, data science, or ML engineering, with 1+ years specifically building LLM-powered agents in production (tool use, structured outputs, memory, multi-step orchestration via LangGraph/Anthropic SDK).

• RAG Expertise:

You have shipped production RAG systems and thoroughly understand grounding, chunking, hallucination control, and refusal triggers.

• Security Mindset:

You treat customer-facing agents as an attack surface and know how to defend against prompt injection and data leakage.

• ML Lifecycle Ownership:

Proven experience with feature engineering, training, serving, and monitoring. Experience in fraud, risk, or abuse detection (imbalanced classes, adversarial users) is a massive plus.

• Statistical Rigor:

Deep knowledge of experiment design, score calibration, and uplift measurement.

• Tech Stack:

Strong Python for backend/services, TypeScript for frontend approval surfaces, and excellent SQL (ClickHouse preferred).

• End-to-End Delivery:

You can spin up lightweight internal apps (e.g., Streamlit, frontend frameworks) so you don't have to wait for another team to make your work visible.

• System Design:

Experience keeping the boundary between probabilistic models and deterministic execution clean, utilizing LLM observability tools (Langfuse, LangSmith, etc.).

Nice to Haves

• Experience in high-trust, transaction-intensive environments (iGaming, FinTech) where fraud prevention and auditability are critical.
• Integrations with CS platforms (Intercom, Zendesk).
• Exposure to blockchain/crypto transaction flows and on-chain data.
• Experience with constrained optimization, bandits, or rule engine/decision-management systems.
• Streaming/event-driven experience (Kafka/MSK consumers, idempotent processing).

How to Apply

We are managing the introduction process for this client exclusively through hiCalibre .

To get introduced with the client via the hiCalibre platform and ensure your application is fast-tracked, create your profile here now: hicalibre.io/join