Hicham Tazi

Leading the end-to-end delivery of AITrust, a groundbreaking, three-component trust framework that embeds real-time guardrails and compliance checks into AXA’s Retrieval-Augmented Generation (RAG) chatbot workflows. Designed first for AXA agents—who demand absolute confidence in AI-driven insights—AITrust ensures every retrieval, generation and response is accurate, policy-compliant and fully auditable, paving the way for a seamless transition to direct-to-customer deployments.

𝐊𝐞𝐲 𝐀𝐜𝐡𝐢𝐞𝐯𝐞𝐦𝐞𝐧𝐭𝐬 & 𝐈𝐦𝐩𝐚𝐜𝐭

• 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬-𝐂𝐫𝐢𝐭𝐢𝐜𝐚𝐥 𝐏𝐢𝐥𝐨𝐭: Spearheaded a secure, enterprise-grade RAG chatbot integration for AXA agents, slashing answer-validation time by 50% and boosting agent confidence.

• 𝐑𝐢𝐬𝐤 & 𝐂𝐨𝐦𝐩𝐥𝐢𝐚𝐧𝐜𝐞 𝐛𝐲 𝐃𝐞𝐬𝐢𝐠𝐧: Enforced continuous checks to prevent misinformation, bias or policy violations—driving an 80% reduction in content-related incidents and satisfying stringent regulatory requirements.

• 𝐑𝐚𝐩𝐢𝐝 𝐓𝐢𝐦𝐞-𝐭𝐨-𝐕𝐚𝐥𝐮𝐞: Delivered a production-ready solution in under 8 weeks, securing executive sponsorship and unlocking a roadmap for customer-facing AI services.

• 𝐌𝐨𝐝𝐮𝐥𝐚𝐫, 𝐒𝐜𝐚𝐥𝐚𝐛𝐥𝐞 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞: Defined a plug-and-play ecosystem—lightweight SDK, independent validator packages, serverless backend—enabling parallel development and zero-downtime updates across multiple lines of business.

• 𝐅𝐮𝐭𝐮𝐫𝐞-𝐑𝐞𝐚𝐝𝐲 𝐆𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞: Established end-to-end audit trails, telemetry and GDPR-aligned data controls, positioning AXA Partners as a leader in responsible, trustworthy AI.

By transforming AI from a “black box” into a governed, auditable asset, I’ve empowered AXA to innovate fearlessly—first inside the agent desktop, next in direct customer engagements—while safeguarding trust, compliance and business value.

As a member of the Core Capabilities team, empower other teams by provisioning robust Big Data and Data Science Frameworks.

𝑯𝒊𝒈𝒉𝒍𝒊𝒈𝒉𝒕𝒔:

• Developed an advanced multi-agent graph knowledge RAG, designed for automatic experimentation and optimization of Spark job configurations to reduce costs. Integrated with Prometheus for monitoring and alerts, alongside a Databricks REST API connection, it doubled as an alert system and an interactive Chatbot.

• Contributed to developing a cloud-agnostic MLOps framework, designed to streamline the training/deployment of ML/DL models with intuitive templates.

• Monitoring ETL processes handling 8 PB of data, leveraging Kafka and Databricks to significantly enhancing data processing efficiency and system reliability.

𝑺𝒕𝒂𝒄𝒌:

Languages: Scala | Python | Java | SQL

Processing: Spark | Databricks

CI/CD: Git | Jenkins

Orchestration: Airflow | Datadog

Deployment: Kubernetes | Azure

Storage: Datalake | PostgreSQL

𝑶𝒃𝒋𝒆𝒄𝒕𝒊𝒗𝒆:

Deliver high-impact PowerBi dashboards to monitor vital KPIs associated with Transilien terminals and CABs validations.

𝑯𝒊𝒈𝒉𝒍𝒊𝒈𝒉𝒕𝒔:

• Processed daily JSON data in Datalake using Spark Scala.

• Created business tables and used Spark for ML insights.

• Engineered a Java application for Job flow .

• Orchestrated robust data pipelines through Airflow and monitored using Datadog.

• Integrated data into PostgreSQL and transformed raw data into actionable insights on PowerBi.