Optimise · Automate · AI · Sustain
I help businesses improve
Operational Efficiency.
From creating mathematical models for solving business problems to automating entire workflows with AI to improve efficiency.
About
Who I Am
Abinauv Selvaraj is a business consultant and technology advisor based in Chennai, India, specializing in operations optimization, AI and machine learning solutions and technology consulting. He holds an MS in Business Analytics from Lancaster University (UK) and a BE in Mechanical Engineering from Anna University (India), with research in mathematical optimization of business problems and a registered Indian design patent.
Trading & Markets
Systematic Trader
Beyond consulting, I actively trade commodity markets using quantitative models I've built and refined over six years.
Risk Optimization
Developed and deployed automated risk optimization models for commodity options trading, incorporating expected value predictions and dynamic risk-reward optimization with defined constraints.
Predictive Algorithms
Built predictive algorithms for commodity price movements (gold, silver, crude oil) using traditional analysis and machine learning techniques, executing hedged futures positions based on model signals.
Consistent Returns
Achieved 20% CAGR over a 6-year period through a systematic trading approach, demonstrating consistent risk-adjusted returns. Continuously refined trading models with hyperparameter fine-tuning of AI models on multiple levels.
Expertise
What I Do
Mathematical Thinking
The Art of Optimization
Interactive 3D visualizations of the optimization techniques I use daily — from simplex pivots on polytopes to simulated annealing on rugged landscapes.
Linear Programming — Simplex Method
A particle walks vertex-to-vertex along the edges of a convex polytope, always improving the objective — the simplex algorithm in action. The sliding plane shows the objective function sweeping toward the optimal corner.
Mixed-Integer Linear Programming
The LP relaxation optimum (amber) sits at a fractional vertex, while the true integer optimum (green) lies at a nearby lattice point. A cutting plane sweeps through, trimming the feasible region — the core of branch-and-cut solvers.
Non-Linear Programming — Gradient Descent
A glowing particle descends Himmelblau's function, navigating a non-convex landscape to find one of four local minima. This is how NLP solvers chase optima when the feasible region or objective is curved.
Heuristics — Simulated Annealing
Forty particles explore a rugged multi-modal landscape with random jumps, accepting worse moves early on. As the temperature cools (bar, bottom-left), they settle near the global minimum — trading optimality guarantees for speed on NP-hard problems.
Track Record
Results That Speak
Ready to optimize your business?
Let’s discuss how I can help you solve complex problems with data-driven solutions.
Get in Touch