Fitahiantsoa Finaritra RAZAKANARY
AI & Automation Project LeadFounder of ORIAPractitioner-researcher in federated learning
MSc Big Data & AI (ESTIA) in progress
Profile
AI & Automation Project Lead for seven companies of a group, and founder and software architect of ORIA, I turn business problems into measurable, secure and well-governed AI solutions — data pipelines, generative AI and RAG, APIs and platforms. Alongside an MSc in Big Data & AI (ESTIA), I conduct applied research on federated learning across heterogeneous companies, applied to probabilistic forecasting of customer collections.
Professional experience
AI & Automation Project Lead — Compagnie MADECASSE
Since October 2026Scope: PRICE · The Land · EMIP · Motostore · CASA · SK Pharma · Madecasse
Leading artificial intelligence and data science projects for seven companies with varied activities, working directly with executive management: some integrations are carried out company by company, others are shared across the group.
Responsibilities
- Defining and steering the group’s AI, data and automation roadmap.
- Analysing business problems and translating them into modelling, experimentation and evaluation objectives.
- Designing and developing solutions: data preparation, choice of approaches, integration and continuous improvement.
- Defining shared architecture, security and data governance principles for the group’s companies.
Ongoing AI & data projects
- Federated forecasting of financial time series: a collaborative training set-up for the cash-flow forecasting of the group’s companies. Models are trained locally and only their parameters are aggregated into a shared model, without centralising financial data, while accounting for heterogeneous (non-IID) distributions and each company’s own temporal dynamics.
- Shared AI architecture: design of an access gateway to AI models and services (API Gateway), with access control, request traceability and usage monitoring across the group.
- Document analysis for recruitment (several group companies): design of information-extraction and document-analysis processing to structure application data and support its assessment.
Areas: federated learning, financial time series, data science, AI architecture, data governance, backend and API development
Founder & Software and AI Architect — ORIA Madagascar
Since 2025Digital transformation & artificial intelligence · Antananarivo
Main activity until September 2026; since October 2026, targeted engagements carried out alongside my current role.
Founder and software architect of ORIA Madagascar, a company specialising in digital solutions, data exploitation and the integration of artificial intelligence to support organisations in their transformation.
Key responsibilities
- Defining the software architecture of applications and platforms.
- Designing and developing AI solutions: LLMs, generative AI, intelligent assistants, process automation.
- Building modern data architectures: collection, processing, analysis and visualisation.
- Developing web applications and scalable SaaS platforms; integrating APIs, cloud services and automation tools.
- Steering technology choices and enforcing good development practices.
- Working with clients to analyse their needs and design high-value solutions.
Engagements and achievements
- Client engagement (confidential) — generative AI on business data: contributed to a Python platform for the intelligent use of business data: multi-source ETL pipelines, a RAG pipeline (Hugging Face embeddings, FAISS vector index), integration and adaptation of Mistral 7B, FastAPI backend services and REST APIs, PostgreSQL storage, evaluation and optimisation of the pipelines.
- Same engagement — multi-region service continuity: an automated system monitoring domain-name availability from several regions (checks every 10 minutes), detecting regional outages and switching automatically to fallback domains, with centralised tracking for incident analysis.
- ORIA platform and AI Lab: architecture of a statically generated bilingual web platform (technical SEO, CSP/HSTS security, zero recurring cost) and of three AI demonstrators running entirely in the browser: WebAssembly OCR for invoice extraction, a BM25 search engine for sourced answers over documents, weak-signal detection.
Tech stack: Python, FastAPI, Django, React, JavaScript, PostgreSQL, Docker, Git, REST APIs, LLMs, generative AI, RAG, Hugging Face, FAISS, PyTorch, Mistral 7B, machine learning, data analytics, Power BI
Python Backend Developer — Oceantrade Madagascar
August 2025 – November 2025Bachelor’s final-year internship
Thesis topic: “Advanced monitoring and intelligent detection of fraud and anomalies in an electric vehicle charging station system”
Main achievements
- Developed backend services with Python/FastAPI and ETL pipelines to collect and transform data from the EV charging stations.
- Orchestrated processing with Apache Airflow and handled inter-service messaging with RabbitMQ.
- Statistical analysis of OCPP data to characterise how the stations operate and are used.
- Used Isolation Forest to detect anomalies and atypical behaviour in operational data.
- Used Random Forest to anticipate failures and guide maintenance interventions.
- Contributed to testing, debugging and deploying the platform.
Areas: Backend · Data engineering · Machine learning · Time series · Distributed architecture
Ongoing research
FedCash — Cross-Silo Federated Learning under Data Heterogeneity: Aggregation Weighting and Local Adaptation
Question: when does federated learning across heterogeneous companies improve collection forecasts?
Status: research in progress — pre-registered protocol, no results published at this stage · finaritra.oria.mg/en/research/
Education
Master 2 — MSc Big Data & Artificial Intelligence
2026–2027ESTIA — taught at IT University, Madagascar — In progress
Machine learning and deep learning, distributed Big Data management, NoSQL and data lakes, relational and object SQL, cloud programming.
Executive Data Manager
2026–2027INSCAE — professional programme — In progress
Data management, administration, modelling and governance.
Certificate in Automation and Artificial Intelligence Integration
2026University of Antananarivo — Obtained
Master 1 in Computer Science
2025–2026IT University, Madagascar — Completed
Machine learning, artificial intelligence, operations research, distributed programming and web services, enterprise databases.
Bachelor’s degree in Software Development
2022–2025IT University, Madagascar — Awarded in 2025
Programming and algorithms, software architecture, databases, vector and matrix calculus, probability and statistics, optimisation.
Scientific baccalaureate — series C (maths & physics)
2022Obtained
Skills
Data science & machine learning
Exploratory and statistical analysis · Data preparation · Feature engineering · Dimensionality reduction (PCA) · Supervised and unsupervised learning · Model evaluation
Tools: Pandas, NumPy, Scikit-learn, PyTorch, Matplotlib, Seaborn
Forecasting & federated learning
Time series · Probabilistic forecasting · Forecast calibration · Federated learning on heterogeneous (non-IID) data — being put into practice
Generative AI
LLMs · RAG · Fine-tuning · AI agents · AI API integration
Tools: LangChain, Hugging Face Transformers, FAISS
Data engineering & automation
ETL/ELT pipelines · Workflow orchestration · Process automation
Tools: Apache Airflow, Kafka, RabbitMQ
Development & architecture
REST APIs · Microservices · API Gateway · SOLID principles
Tools: Python, FastAPI, Django, Java, Spring Boot, Hibernate, Jakarta EE
Data & governance
Data governance · Personal data protection (Malagasy Law No. 2014-038)
Tools: PostgreSQL, Oracle Database
Languages
Malagasy: native · French: DELF B2 (June 2025) · English: working proficiency
Interests
Music (piano, guitar, saxophone), swimming