Harmoniq Intelligence.
A boutique Operations Research and Applied AI consultancy. We engineer custom optimisation models, predictive pipelines, and scalable data architectures for complex enterprise problems, turning operational bottlenecks into measurable ROI.
Founded by an applied mathematician and a senior data architect. We combine mathematical rigour with production-grade data engineering.
Two pillars of engineering excellence
Deep technical expertise across mathematical optimisation and applied AI, delivered through embedded contract engagements.
Constraint Programming & MIP Solvers
Formulation and solution of complex scheduling, assignment, and allocation problems using SCIP, Gurobi, OR-Tools, and CPLEX.
Linear & Integer Programming
Exact and heuristic methods for large-scale optimisation across supply chain, logistics, and resource planning.
Decomposition Strategies
Benders decomposition, column generation, and custom cut-callback architectures that make monolithic models tractable.
Routing, Scheduling & Resource Allocation
Vehicle routing (CVRP/VRPTW), workforce rostering, and multi-resource timetabling algorithms to minimise fleet idle time, slash fuel costs, and maximise asset utilisation.
Machine Learning & Predictive Models
Gradient-boosted models, time-series forecasting, and classification pipelines for demand planning and anomaly detection, ensuring you never overstock inventory or suffer unexpected machine downtime.
LLM Orchestration & AI Agents
Retrieval-augmented generation, agent chains, and structured output parsing for enterprise workflows.
Data Pipelines & Integration
End-to-end ETL, SQL and graph databases (Neo4j), data validation, and automated reporting systems.
Full-Stack Applications
Production dashboards, internal tools, and client-facing applications built with React, Next.js, and TypeScript.
26x Solver Speedup
Before launching Harmoniq Intelligence, our founder engineered a custom Logic-Based Benders Decomposition (LBBD) architecture for a live enterprise ERP platform. By decoupling monolithic assignment logic and routing the master problem through a SCIP solver, model scale was increased by 4x while complex scheduling runtimes were reduced from 26 minutes to under 60 seconds.
This sub-minute runtime enabled dynamic, intra-day re-routing instead of relying on stale overnight batch processing.
Have a complex optimisation challenge?
We scope engagements in days, not months. Share your problem and we'll outline a technical approach - no commitment required.