Walter Sebastian Gisler Mathematics & optimization
for real-life problems

The Mathemagician

I use mathematics and computer science to magically solve real-world problems.

Fig. 1 — Travelling salesman A salesman visits 48 cities. A first route is laid down, then two local searches improve it: 2-opt uncrosses pairs of edges and Or-opt moves runs of up to three cities somewhere cheaper. Neither can promise the shortest route, so each stage is measured against the best of many restarts. Click to add a city, drag one to move it.
Length — —
01

Services

The problems worth bringing to me are the ones where the mathematics is the hard part. I solve that part, then build software that makes the answer usable.

  1. Scheduling & optimization

    Custom software for scheduling and optimization problems too complex for off-the-shelf tools.

    Leagues and tournaments, staff rosters, timetables, venue and resource allocation, routing. When the rules multiply and the constraints start to conflict, I model the problem precisely and build a solver around it. Exact methods such as mixed integer and constraint programming2 are combined with purpose-built heuristics1 where exact methods stop scaling. The result plugs into your existing systems, produces plans people trust, and can explain why a schedule looks the way it does3.

    • Mixed integer programming
    • Constraint programming
    • Metaheuristics
    • Large neighbourhood search
    • Decomposition methods
  2. Simulation

    See a decision play out thousands of times before you make it once.

    Monte Carlo methods4 for risk, pricing and forecasting. Discrete-event simulation for queues, operations and capacity planning. Agent-based models where the behaviour you care about emerges from many interacting parts. Each model is calibrated against real data and reports its uncertainty honestly, with confidence intervals, sensitivity and scenario analysis. Variance reduction keeps the runs short enough to be useful, and simulation-based optimization searches for the policy that holds up best across all those futures.

    • Monte Carlo
    • Discrete-event simulation
    • Agent-based models
    • Scenario & sensitivity analysis
    • Variance reduction
    • Simulation-based optimization
  3. AI & agentic engineering

    Helping companies use AI where it actually pays off.

    I start by separating what AI does reliably today from what only demos well, then help you build the parts that hold up. That means agents that plan, call tools and act on real systems, with the evaluations and guardrails that tell you when they are right. For engineering teams, it means getting productive with coding agents: workflows, review practices, and the codebase conventions that let agents do their best work.

    • AI strategy
    • Agent design & tool use
    • Evaluation & guardrails
    • Coding agents
    • Team enablement
  4. Software development

    Well-engineered software, from first prototype to production.

    Backends and APIs, data pipelines, algorithmic services and interactive visualizations like the figures on this page. The code is tested, documented and readable, so the people who inherit it can maintain it.

    • Architecture
    • Backends & APIs
    • Data pipelines
    • Interactive visualization
    • Performance tuning
  5. Sports organizations

    Analytics, AI and optimization for clubs, leagues and federations.

    Sport is full of hard problems. Fixtures have to be scheduled around venues, travel and fairness3. Referees and officials need assigning, performance needs measuring, and results invite ratings, rankings and forecasts. I help sports organizations find where data, AI and optimization make a measurable difference, then build the tools that deliver it.

    • Fixture & tournament scheduling
    • Referee assignment
    • Performance analytics
    • Ratings & forecasting
    • AI adoption
02

Areas of interest

Optimization
Linear & mixed integer programming2 Combinatorial optimization1 Scheduling problems3 Heuristics1
Mathematics
Mathematical modelling Discrete mathematics Statistics & probability4
Computing
Algorithm development High performance computing
Data
Data visualization Machine learning5
03

Worked examples

Most real problems turn into one of a few shapes once you write them down properly. Four more of them follow, small enough to play with.

Fig. 2 — Linear program Every constraint cuts the plane in half, and what survives is a convex polygon. Push the objective's level lines as far as they go and they leave at a corner, which is why the simplex method only visits corners. Ask for whole numbers and the best point hides inside, on the integer hull (dashed). Closing that gap is what branch-and-bound is for. Drag or use the arrow keys to aim the objective.
—
Fig. 3 — Round robin The circle method. Fix one team in the middle and seat the rest around a circle: each round's matches are parallel chords. Rotate by one seat for the next round, until every pair has met exactly once. The table records the round in which each pair plays.
— Teams 8
Fig. 4 — Buffon's needle Drop needles as long as the gap between floorboards and each one lands across a crack with probability 2/π. Count the crossings and you can estimate π by throwing things on the floor. Monte Carlo error only shrinks like 1/√n, so every extra digit costs a hundred times more needles.
—
Fig. 5 — k-means Lloyd's algorithm: tie every point to its nearest centre, move each centre to the mean of its points, and repeat until nothing changes. It always stops, but only at a local optimum. Reseed and the same data can settle into a different, sometimes worse, answer. The dashed lines are the paths the centres took. Click to drop a handful of points.
— k 4
04

Other things I like

Cooking
And, consequently, eating.
Sardines
The gourmet kind, in a nice tin.
Sports
Skiing, swimming, running, mountain biking, hiking.
Greece
The place, the food, the sea.
Macroeconomics
As a spectator sport.
05

Contact

Have a problem that needs some mathematics? Tell me about it.

Your details are only used to reply to you.