Decisions that add up.
We build the systems that get you there.
Haering Solutions is a Swiss consultancy with roots in operations research and decision intelligence. We build the software businesses use to make their hardest calls: pricing, scheduling, routing, energy bidding, network design, policy, internal tooling. Sometimes that means a heavy-duty optimization solver. Sometimes it means a lighter tool with AI doing parts of the work. We pick what fits the problem, and we ship where the right answer is worth millions and the wrong one is hard to undo.
Trusted by leaders across energy, mobility, finance & public sector
We start with the decision.
The shape of an engagement depends on the problem. A pricing engine can need months of solver tuning, while a smaller decision tool might only need a clean forecast and a sensible interface. We are comfortable doing either, and most engagements end up somewhere in between.
On interpretability
Clients eventually ask the same question: why did the system pick this answer, and how good is it? Most ML systems cannot really say. With the optimization we build, you can. Every constraint is explicit, every objective is stated in math, and the gap to optimal comes with a number attached.
- 01
Mathematical optimization, end-to-end
This is the deep end. We build solvers for the hard problems: pricing, routing, scheduling, energy bidding, network design. The work spans mixed-integer, non-linear, stochastic, and robust formulations, with proper algorithmic engineering when the off-the-shelf tools fall short. PhD-grade operations research, delivered as production code.
- 02
Algorithmic R&D
When the off-the-shelf solver gives up, we write the algorithm ourselves. Branch-and-bound, decomposition methods, metaheuristics, problem-specific tricks. A lot of what we do here has ended up published, but the goal is always to make the runtime work in production.
- 03
Forecasting & predictive modeling
The signal layer that feeds the optimization. Discrete-choice models, time-series forecasting, demand simulation, classification, all calibrated on your data rather than someone else's public benchmark.
- 04
Decision-support tools, AI-augmented
Models only matter if your team can use them. We build the APIs, dashboards, and integration layers around the model. Where it earns its keep we add AI on the surface: document parsing, natural-language queries, summarization of long inputs. We work through the major frontier providers via API rather than running our own LLM stack.
Three things we mean when we say AI.
AI is a broad word, and it covers genuinely different things. We use generative models when the input is unstructured, like text or documents. We use predictive models to forecast and classify. We use mathematical optimization when the decision has to be provably good. Most engagements use at least two of the three.
The AI that decides.
Our deepest specialty. We model the business decision as a mathematical optimization problem and solve it. The output is a recommendation with hard guarantees: a feasibility certificate, an explicit list of constraints and objectives, and a number telling you how close to optimal the answer is. Most of what we ship in this lane has been published in peer-reviewed OR journals before it hit production.
When the decision has constraints, stakes, and a quality bar.
From data to signal.
Forecasting, demand modeling, discrete choice, classification, signal extraction. We pick the model that fits your data, calibrate it on the actual problem, and treat its output as the input layer to whatever decision engine sits downstream. The right model is often a smaller one than the trends would suggest.
When you need to know what is likely, before deciding what to do.
Unstructured in, structured out.
We integrate large language models where they earn their keep: document parsing, natural-language interfaces over structured systems, knowledge-work automation, agentic glue between systems. We work through the major frontier providers via API. Where a simpler tool would do the job, we say so.
Document-heavy tasks, natural-language interfaces over structured systems.
You do not always need an LLM, or a solver, or even a sophisticated model. Sometimes the right answer is a careful heuristic in a spreadsheet, and we will say so when it is. The objective of an engagement is to make the decision better, not to maximize the technology stack.
A glimpse of recent engagements.
Work across energy, mobility, finance, public health, and travel. Senior staff lead every engagement.
A Custom Portfolio Optimizer, Up to 40% More Risk-Return Efficient Than Standard Heuristics
We built swissQuant a custom portfolio optimizer for risk-contribution-constrained allocation, a non-convex problem standard MIP solvers cannot solve to certified optimality. On real five-year MSCI World data, the heuristic alternative is up to 40% less risk-return efficient than our optimizer on 33-asset portfolios; in the most non-convex regime it drifts so far it returns a positive objective where the true optimum is negative.
Day-Ahead Bidding That Adds €0.9M of Realized Profit on a 900 MW Pumped-Storage Plant
We co-developed the optimization framework behind Alpiq's day-ahead bidding for the 900 MW Nant de Drance pumped-storage plant. Across 4 months of real cleared prices (Jan, Apr, Jun, Oct 2022) the stochastic version adds €0.9M (+3.2%) of realized profit over the deterministic approach Alpiq's traders use today. Enable the CVaR risk knob and the same engine trades ~2.5% of expected profit for ~70% lower tail risk.
A City-Scale Activity Scheduler That Reproduces the Swiss Microcensus on an Open-Source Stack
We co-authored the activity-based scheduling framework that turns the temporal layer of SBB's travel-demand work from rule-based heuristics into a real optimization problem. On a 46,970-schedule synthetic population of Lausanne workers, the simulated activity profiles reproduce every empirical signature of the Swiss microcensus, morning peak, lunch dip, evening leisure ramp, without any rule-based patching.
68 Behavioral Parameters of Swiss Worker Schedule Flexibility
We co-authored the open-access journal paper that estimates, from 10,110 real Swiss microcensus schedules, how flexible Swiss workers actually are about lunch, work timing, leisure, shopping, and the home day. All 68 behavioral parameters land statistically significant at 5%; the resulting calibration plugs into any activity-based travel-demand model. One memorable finding: workers are roughly 350× more averse, per hour, to a short lunch than to a short workday.
A Live Itinerary Engine That Picks the Best 3-Day Tour Out of 200M+ Google Places in Under 90 Seconds
We architected and shipped the optimization engine that turns 200M+ Google Places and a traveler's preferences into a personalized 3-day itinerary in under 90 seconds end-to-end. The optimizer closes a 1163% initial gap to under 10% inside that window. Engine, Google Places ingestion pipeline, and reference UI all live in Unki's traveler-facing app — benchmarked across Geneva, Paris, London and Lausanne.
Targeted Pandemic Restrictions That Cut ~22% of Infections at the Same Economic Cost
We delivered a 15-policy Pareto frontier of pandemic restrictions for the Vaud canton's 814,000-agent synthetic population, each policy paired with its full 60-day simulated epidemic trajectory. The headline finding: smarter age targeting cuts roughly 22% of infections at the same economic cost (72M vs 69M CHF) — restrict the 30–39 work cohort and you control the spread; restrict the 40–49 cohort and infections rise by 100,000 over two months. That kind of age-resolved decision is exactly what a coarse SEIR model cannot deliver, and exactly what a public-health response needs.
The deep end of the toolkit.
Each card below is a methodological capability we have led, co-authored, or supervised. Some has been published in a peer-reviewed journal, others written up as a technical report. Most improved on the prior state of the art by orders of magnitude. When the decision has to be provably good, this is the research bench we draw on.
Exact Choice-Based Pricing, Up to 1.67 Million Times Faster Than the MILP Baseline
We developed three exact algorithms for continuous pricing under any advanced discrete choice model (mixed logit, nested logit, probit, anything with linear-in-price utility): a spatial Branch-and-Bound, a Branch-and-Benders Decomposition variant, and a polynomial-time Breakpoint Exact Algorithm. On the standard parking-pricing benchmark with mixed-logit demand, the BEA solves 1,000,000 demand scenarios in 77 seconds — a 1.67-million-× speedup over Gurobi's MILP and a 300× speedup over CoBiT, the previous ML-specific state-of-the-art. Published in OR Spectrum.
A Pricing Heuristic With <0.2% Optimality Gap and Up to 37 Million × Speedup Over the State-of-the-Art
We introduced the Breakpoint Heuristic Algorithm and its Iterated Local Search extension: a fast, scalable heuristic for the capacitated choice-based pricing problem that delivers sub-0.2% average optimality gap on benchmarks where every exact method times out. Versus the previous capacitated state-of-the-art (LAG), it's at least 79× faster. Versus the previous mixed-logit-specific state-of-the-art (CoBiT), it's up to 37 million × faster — while the ILS extension matches every known optimum. Published open-access in Computers & Operations Research.
BHAMSLE: +17.2% Log-Likelihood Lift on London Mode-Choice Mixture, in Half the Runtime of Multistart Biogeme
We took the Breakpoint Heuristic Algorithm developed in our OR Spectrum pricing work and adapted it to Maximum Simulated Likelihood Estimation of advanced discrete-choice models with latent classes and continuous mixtures. On the London mode-choice discrete-continuous mixture, BHAMSLE-initialized Biogeme reaches a 17.2% better log-likelihood than default initialization (LL = −243.68 vs −294.38), in roughly half the wall-clock time that Multistart Biogeme with 20 restarts needs to match the same quality. Submitted to Elsevier.
A Decentralized Facility-Location Optimizer That Solves 11× More Scenarios Than the Textbook MILP
We supervised the design of a decentralized capacitated facility-location optimizer that integrates advanced discrete-choice demand directly into the MILP, with capacity handled by exogenous priority queues. On the real Dresden Gymnasium-location dataset, the Progressive Scenario Expansion method solves instances of 14 schools × 558 students × 110 scenarios within a 24-hour limit — where the textbook MILP stalls beyond 10 scenarios. With the full feature stack (priority valid inequalities + complete warm start + extended-entropy ordering), PSE beats its basic configuration by up to 95% and the RAW MILP by 28% on cases both can solve.
Electric Autonomous Ride-Sharing Optimization, 31× Faster Than the Published Branch-and-Price Baseline With Up to 11× Tighter Gaps
We supervised the design of an event-based mixed-integer formulation for the electric autonomous Dial-A-Ride Problem that beats the published Branch-and-Price baseline of Su (2023) by 31× on average runtime and 11× on average optimality gap on hard instances. The companion Adaptive Large Neighborhood Search heuristic provides a deployment lane: 10× faster than the exact method at less than 1% loss in objective value. On the largest benchmark instance (r8-96, 96 users, 8 vehicles), ALNS finds a better solution than the exact method in 9× less time.
Joint Mode-and-Destination Choice Sets, Now Live in EPFL's OASIS Pipeline
We co-authored the first choice-set generator that handles the mode/destination coupling jointly and ships a closed-form bias-correction term any activity-based-model estimator can plug into. Now the choice-set-generation building block of the OASIS activity-based modeling pipeline at EPFL Transp-OR.
Constraint Programming, 750× Faster Than MILP on Activity-Based Scheduling
We co-supervised a head-to-head benchmark of constraint programming versus the dominant MILP formulation for activity-based scheduling, and built CP variants that solve the problem 750× faster on 5-activity instances. The deeper lesson: not every optimization problem deserves the same hammer.
A senior team led by Dr. Tom Haering.
We are a consultancy with roots in operations research and decision intelligence, founded by Tom Haering after his PhD at EPFL. We keep the team deliberately lean — senior staff lead every engagement, and nothing gets handed off to a junior.
Dr. Haering completed his PhD in 2025 at EPFL's Transport and Mobility Laboratory under Prof. Michel Bierlaire, working at the intersection of discrete choice modeling, mixed-integer optimization, and decision-aided algorithms. His work has been published in OR Spectrum, Computers & Operations Research, and Transportation (Springer).
Before that, a Master's in Mathematics from ETH Zurich (Institute for Operations Research), with a thesis on portfolio optimization done in cooperation with swissQuant. Before that, an undergraduate background in mathematics and a fascination with building things in code that never went away.
He has supervised numerous master's and semester projects in optimization and transport modeling, taught decision-aid methodologies, and reviewed for international transportation journals. The point: when we say senior leadership, we mean it.
Education
PhD, EPFL · Transport & Mobility Laboratory (2025). MSc Mathematics, ETH Zürich · Institute for Operations Research.
Publications
Peer-reviewed in OR Spectrum, Computers & Operations Research, Transportation. 7+ technical reports and conference papers.
Teaching
EPFL teaching assistant in Decision-Aid Methodologies, Optimization, Mathematical Modeling of Behavior, and Discrete Choice.
Languages
English · German · French · Python · Gurobi · CP-SAT · Pyomo · OR-Tools · Biogeme.
Where
Headquartered between Lausanne and Zurich. Engagements globally, remote-first.
Style
We design around the actual decision. We deliver code that runs in production, not slides.
Let's start optimizing
Tell us the decision you want to optimize. We'll send a concrete plan back.
We work with engagements ranging from focused two-week sprints to multi-quarter optimization platforms.
Most of our clients arrive with a vague idea of what they want ("we need better pricing", "we need smarter routing"). We help them turn it into a tractable, mathematically defensible problem.
mail@haeringsolutions.ch