Your first deliverable is a production SQL BI analyst agent: a Slack-native agent that answers business questions with governed SQL over the analytical warehouse, with validated queries, sanity-checked results, and cited evidence behind every number. From there, the role balances four crafts in roughly equal measure: AI agents, machine learning, data science, and the dashboards and surfaces that expose them.
• Build and own analyst agents end to end: Slack-native agents that translate natural-language business questions into governed SQL over the analytical warehouse, answer executive P&L questions, post daily health briefings, and explain metric movements with data-backed root cause analysis; guarded by SQL and schema validation, result sanity checks, and cross-checks of daily metrics against canonical reporting views.
• Extend into customer-facing agents: intent triage and routing, retrieval-grounded (RAG) responses over versioned knowledge bases with strict abort-and-escalate fallbacks, multi-turn conversational state machines, localized brand voice, and escalation logic; integrated with our helpdesk and CRM platforms (webhook ingestion, session lifecycle management, intent metadata tagging, automated escalation tickets with pre-packaged tool context).
• Build the risk-stratified tool layer between agents and back-office APIs: read-only context-gathering tools, information-first validation before any action, multi-turn confirmation workflows, and a per-tool switch that moves low-risk mutations from lead approval to fully autonomous execution as evidence accumulates. Harden agents against adversarial input: prompt injection screening, confidence-threshold freezes on sensitive intents, silent security escalation paths, and defences against tool misuse and data exfiltration.
• Build agents up the autonomy ladder using LangGraph, the Anthropic Agent SDK / Model Context Protocol (MCP), or equivalent orchestration frameworks. Engineer the closed feedback loop (corrections capture, proven-query and semantic memory), decision audit logging, and evaluation harnesses, including regression suites proving new tools or intents introduce zero degradation to existing paths.
• Build and productionise the models behind the platform’s decision signals: churn, lifetime value, and bonus-sensitivity models for engagement; composite player risk scores across identity, payment, gameplay, bonus, and network signals; collusion, bot-play, and multi-accounting detection; and anomaly detection for treasury and payments.
• Ship models as governed signals, not notebooks: versioned, SLA’d contracts with the decision engine and your agents, with freshness, drift, and calibration monitoring and automated retraining paths, served across real-time (Kafka/MSK), near-real-time, and batch (ClickHouse) tiers.
• Own multi-vector withdrawal risk scoring: per-vector scores with cited rationale and confidence, evidence-aware aggregation, and automatic re-scoring when late evidence lands.
• Codify business rules with domain owners and keep policy auditable: translate policy into deterministic, configurable rules; simulate and backtest every rule or threshold change against historical data before activation; design holdouts and control groups to measure true uplift; and run the deep-dive analyses that feed both your agents and the executive team.
• Build the supervisor and approval surfaces for your agents: review queues with one-click action proposal cards for high-risk mutations, searchable session replay exposing prompts, model outputs, reasoning chains, and tool calls, and a structured grading module whose output feeds evaluation and fine-tuning datasets.
• Design and ship the dashboards through which the business consumes your work: decision audit views, agent performance dashboards (correction rate, failure rate, decision volume by rule and vector), risk review queues, and KPI views built with the BI team, moving dashboarding toward AI-assisted anomaly detection and explanation.