Dewan Fayzur Rahman

Principal Architect – Data & Analytics at DevScope in Porto, Portugal.

Email  /  LinkedIn  /  Scholar  /  GitHub

Profile

I architect and deploy enterprise data platforms, production Generative AI pipelines, and mathematical optimization systems. With over 15 years of technical experience spanning scientific research at INESC TEC and enterprise consulting at DevScope, I specialize in translating complex business constraints into scalable, deterministic systems.

Core Architectural Focus:

• Generative AI & Retrieval: Enterprise Graph-RAG, Knowledge Graphs, Document AI platforms, and Speech AI pipelines with strict PII anonymization layers.
• Modern Lakehouse & MLOps: Resilient data foundations using Databricks (Delta Lake, Unity Catalog, Spark) and Microsoft Azure for batch, streaming, and model serving.
• Mathematical Optimization: Formulating Mixed-Integer Linear Programming (MILP) and metaheuristics (CPLEX, Gurobi, AMPL, GLPK) for energy scheduling, supply chain, and workforce planning.

Distinctions & Awards

• 1st Prize: Animal Behavior Challenge (ABC2018, Kyoto, Japan) – International bio-logging trajectory classification challenge organized by Kyoto University. Stacked ensemble architecture published on arXiv (1808.08613).
• 1st Prize: Kaggle Days Porto (2019) – Talkdesk WFM Challenge for contact center staffing demand forecasting under strict competition constraints.
• Enterprise AI Industry Recognition: Technical leadership behind global workforce AI platforms featured in major business media (Forbes).
• INESC TEC Limelight Distinction: Probabilistic renewable power generation forecasting modules integrated into national electric grid operations.

Professional Experience

DevScope  |  Porto, Portugal

Principal Architect – Data & Analytics (Jan 2026 – Present)
Data Science Technical Lead (Jan 2025 – Dec 2025)
Data Scientist (Oct 2017 – Dec 2024)

• Lead technical architecture, data governance frameworks, and system design reviews across enterprise AI and advanced analytics programs.
• Speech AI & Post-Call Analytics: Architected and led delivery of high-volume speech analytics platforms using Databricks Unity Catalog, Azure, Whisper, and LLMs with strict PII anonymization layers.
• Enterprise Workforce AI: Built time-series revenue/payroll forecasting pipelines; architected generative AI recruiter engines saving ~200 hours/recruiter annually; operationalized Databricks MLOps with automated CI/CD and observability.
• Edge Computer Vision & Supply Chain: Engineered GPU edge analytics (YOLO, Azure Batch Shipyard) for hypermarket tracking, and built inventory coverage models (LightGBM, XGBoost) for multinational retail groups.

INESC TEC  |  Porto, Portugal

Researcher (Machine Learning & Forecasting) (Jul 2013 – Sep 2017)
Researcher (Optimization) (Dec 2010 – Jun 2013)

• Developed multivariate statistical and computational intelligence models for short-term renewable energy prediction (wind, solar, hydro) across FP7 European Commission consortia (SuSTAINABLE, evolvDSO, SENSIBLE).
• Formulated Mixed-Integer Linear Programming (MILP) models and matheuristics for power production coordination under the FCT-funded project COORDINATOR using CPLEX, Gurobi, and AMPL.
• Engineered hybrid metaheuristics (PSO, Genetic Algorithms) for energy production planning and non-convex economic load dispatch.

Publications