About Smart Bricks

Smart Bricks is a Dubai-based proptech company building real estate intelligence products powered by high-quality data, analytics, and AI. Our systems process large-scale property listings, transactions, and market signals to generate insights, valuations, and search experiences.

We’re looking for a ML who can build reliable pipelines, enforce data quality, and enable AI/ML teams with clean, versioned datasets.

Roles and Responsibilities

• Build and maintain Automated Valuation Models (AVMs) for real estate pricing
• Continuously improve models by incorporating new datasets and ensuring models remain accurate over time
• Develop predictive models for:: price projections, rental yield forecasts, ROI / appreciation trends, property scoring and ranking
• Design and tune feature engineering pipelines to improve model performance and confidence
• Identify data gaps and define what new data is needed to improve model accuracy (and work with data teams to acquire it)
• Validate, test, and benchmark models using strong evaluation practices
• Monitor model performance and drift (data drift + concept drift) and trigger retraining strategies
• Collaborate closely with backend + data engineering teams to productionize models
• Build model scoring services and assist with integration into real-time APIs
• Work with modern AI workflows including Agentic AI / OpenAI Agents to enhance automation and intelligence in internal systems

Required Skills & Experience

• 6+ years of experience in developing AL Models

Nice-to-Have Skills

• Experience with time-series forecasting models (Prophet, ARIMA, XGBoost forecasting)
• Experience with model monitoring + retraining pipelines
• Familiarity with ML tooling like MLflow, DVC, Weights & Biases
• Experience deploying ML models via APIs (FastAPI, Flask, Docker)
• Familiarity with geospatial datasets and location-based modeling
• Experience with LLM workflows, prompt engineering, and Agentic AI frameworks (OpenAI Agents, LangChain, etc.)
• Knowledge of ranking systems and scoring frameworks

What We Value

• Strong sense of ownership: you treat models like products, not experiments
• Obsession with model accuracy, reliability, and explainability
• Comfort working with messy real-world data
• Curiosity to improve datasets and uncover better predictive signals
• Ability to debug models and pipelines end-to-end (data → features → model → output)

What Success Looks Like

• AVM models that stay accurate and robust across new market conditions
• Improved feature sets and smarter dataset design over time
• Strong monitoring of model drift and confidence degradation
• Well-tested, reproducible ML pipelines with documented assumptions and metrics