Raphael Luta

We provide technical consulting services: – software, data or networking audits – software systems design and implementation – datalake design and implementation, data engineering, data/ML ops – cloud costs optimisation – software team leadership and coaching

Tasked with improving operational efficiency and reliability of offer engine API of Accor CRS (TARS) : – rewrote the data synchronization engine with the central TARS database – significantly improved the core rule engine of the system – simplifier/refactored some core functional components such as price and tax calculations – optimized the data flows with downstream consumers for better end-to-end efficiency of price distribution After 1 year of production of updated system: – 95% less issues opened by customers – x15 increase in throughtput per API server – -70% p50 latency and -50% p99 latency API service time – reduced AWS costs by $ 700.000 / year – avoided 15t of CO2 emissions / year

Design and implement migration plan for DataManager and Cyberguard platforms out of legacy Hortonworks 2.6 Hadoop platform. – Set-up Kubernetes for running Spark jobs – Rework legacy Hive and Spark 1.x ETL jobs into Spark streaming jobs deployed on Kubernetes – Implement a compatibility API to emulate OpenTSDB calls with a Clickhouse backend – Migrate the legacy HBase and OpenTSDB data into a new distributed ClickHouse cluster – Interconnect Spark streaming ETL and clickhouse through Kafka topics – Update the Sendika APIs to query Clickhouse rather than HBase/OpenTSDB – Rework the DataManager front-end application in Vue.js with a DuckDB embedded query engine for customer analytics => hardware requirements divided by 2, ETL latency decreased from several hours to less than 1 minute

As part of the central core data lake team: – improve stability and performance of the existing datalake services (-70% average job latency) and reduce operating and AWS cloud costs (-50% monthly costs) – coach the development team on better software and devops practices. – Rework the CI and CD pipelines and development flows to improve delivery reliability and frequency – develop internal tools to simplify and/or automate operational tasks – support project data engineering teams on Spark job design and optimization and EMR cluster optimization As a separate project team: – design and implement a cost-effective worldwide, near-realtime warehouse service for company-wide SCADA data from industrial plants to be used as backend for business intelligence, predictive maintenance and process optimization applications

In the AI Factory team of Kering Digital: – I helped set up and manage an AWS work environment for the data science and data analytics teams that optimize for developer/scientist productivity and autonomy – I also acted as a software and data engineering support/coach for the teams, especially for projects going into production to ensure code quality, scalability and reliability in production

I designed and implemented the real-time data processing infrastructure for the Marlink Cyberguard maritime cyber defense product (Spark streaming and structured streaming, Kafka and Vert.X)

I lead the development team (6 dev) for a new product for Renault controlling department, analyzing financial impact of model and component mix evolution across all markets and vehicle families. Within that team, my technical roles were mainly overall software and system design, data engineering and processing (Spark/Scala) and resident devops in charge of the product cloud infrastructure (dev and prod) and build pipeline (AWS, Gitlab CI, Packer, Terraform). Additionnally, in the early days of the company, I've also helped on the following subjects: – Job definition and key skill requirements for technical positions – Overall cloud hosting infrastructure and devops pipeline – Help other lead devs bootstrap their projects Later, I've tried to improve and solve common challenges across projects: – User authentication and authorization, with reusable OpenID Connect servers and clients – Hadoop/Spark data engineering practices and tools – Data streams, change data capture and streaming infrastructure

I teach the "Network and Telecom Security" course in the Master 2 Electronic Commerce - Cybersecurity minor

Replace a legacy customer usage tracking reporting solution for Marlink satellite network with a new Hadoop based solution: – Install, automate and validate hardware, OS and hadoop stack (foreman, puppet and ambari) – Set up redundant, high-availability data collection flows from backbone equipments and ensure data quality (flume, kite, kafka) – Define and implement data model and aggregation tasks to correlate and extract customer data in batch and real-time mode (spark, spark streaming, opentsdb, hbase) – Define and implement a REST access API and streaming API for customer portals to access data – Build first customer application portal (vert.x, react.js)

I designed and implemented a scalable TR-69 device management system for the operator next generation set top boxes and DSL modems, based on a reactive service architecture (Vert.x) and distributed database (Cassandra). I also contributed to the design and prototype a new unified service hosting infrastructure based on Mesos and Docker and mentored other team members to leverage all these new tools for their own projects.

In the customer self-care IT team, my job was to provide technical expertise and support to the marketing and development teams, especially on all performance and security matters. Additionally, I lobbied for, designed and ran an internal big data analytics platform (built on Hadoop, Elasticsearch and MongoDB) that helped us improve user experience, diagnose technical incidents and identify fraud patterns across all our web properties.

Within the product marketing group, lead a small development team (3-5 devs + 1 PM) capable of producing "Quick & Clean" innovative services that could be inexpensively productized and marketed. Key goals for the team were development velocity and cost-efficient operations. We delivered many production services during that period from the complete SFR Facebook services package to a working IMS AS delivering smart call routing services.

For this Internet music streaming/social locker start-up, I took on the usual hands-on CTO role: product architecture and development, hiring and building the technical team and being the main technical contact for all our partners (music majors, collecting societies, etc...) We were the first streaming music startup to secure a licence from the majors without a DRM system.

Mediaserv is a French Telecom Operator in the Caribbean and Reunion Island. I designed and implemented the OSS infrastructure (mainly AAA and CPE management services) for their triple-play service.

Service Design, specifications, service integration and support/administration for the white label, turn key triple play solution to be sold through major French retail brands such as Darty or FNAC.

Main developer for dynamic websites built by the company (e.g. Degremont, Lyonnaise des Eaux, Voyageurs du Monde, Interflora, etc...) Professional Trainer on major Internet related topics: Java, Internet networking, Unix administration, etc...