Shell.ai Hackathon for Sustainable and Affordable Energy 2025
Shell.ai Hackathon for Sustainable and Affordable Energy brings together brilliant minds passionate about digital solutions and AI, to tackle real energy challenges and help build a lower-carbon world where everyone can access and afford energy.
Click on the link above to register your interest to participate in the Shell.ai Hackathon 2025. When you click on “Sign Up Now†you’ll be asked to provide your details which we will use to send you the information on how to join the Hackathon on July 4, 2025.
Fuel Blend Properties Prediction Challenge
In the sixth edition of the Shell.ai Hackathon for Sustainable and Affordable Energy, you will immerse yourselves in the critical field of fuel blending. In the previous five editions, we addressed some of the digital challenges around the energy transition: windfarm layout optimisation (2020), irradiance forecasting for solar power generation (2021), optimal placement of electric vehicle (EV) charging stations (2022), supply chain optimisation for biorefineries (2023), and fleet decarbonisation (2024). This year, we focus on blend properties estimation for sustainable fuel.
In this hackathon, your challenge is to develop models capable of predicting the final properties of complex fuel blends based on their constituent components and proportions. The endgame is to engineer powerful predictive tools that can guide the industry in formulating the next generation of sustainable fuels, potentially accelerating the transition to a net-zero future, without compromising on model excellence.
Disclaimer: Please note that for the purposes of this challenge, the term “sustainable fuels†encompasses fuels derived from renewable or low-carbon sources; such as bio-based feedstocks, synthetic fuels generated using renewable electricity (e-fuels), and fuels produced from waste materials. These alternatives are intended to significantly reduce lifecycle greenhouse gas emissions when compared to conventional fossil fuels.
Cautionary note: /investors/disclaimer-and-cautionary-note.html
Eligibility and Prizes
General Edition
Eligibility:
Open to all individuals and teams of up to 4 members (except Shell employees and contractors)
Shortlist and winners:
- Level 1 - The participants are invited to develop mathematical models to optimise fleet decarbonisation strategies. 20 Participants with top scores in the private leaderboard (regardless of the edition) will be shortlisted for Level 2.
- Level 2 - Shortlisted teams will develop a prototype, based on their solution. The top three finalist will be announced based on the judging criteria.
- In Level 3, the finalists will pitch their solutions live and the winners will be selected.
Prizes:
- 1st prize: 2,500 USD
- 2nd prize: 2,200 USD
- 3rd prize: 2,000 USD
Special University Edition
Eligibility:
Open to anyone currently studying at or associated with any university/institute/college (students, PhD, Postdoc, Technical Assistant, Research Assistant, faculty, etc.)
Shortlist and winners:
- Level 1 - The participants are invited to develop mathematical models to optimise fleet decarbonisation strategies. 20 Participants with top scores in the private leaderboard (regardless of the edition) will be shortlisted for Level 2.
- In Level 2 - shortlisted teams will develop a prototype, based on their solution. The top two finalists will be announced based on the judging criteria.
- In Level 3 - the finalists will pitch their solutions live and the winners will be selected.
Prizes:
- University Edition Winner: 1,500 USD
- University Edition runner up: 1,000 USD
- Additionally, for both teams, an opportunity to start External Technical Collaboration between Shell and the institute that the winner is part of, subject to Shell’s criteria and regulations.
Special Start-up Edition
Eligibility:
Open to registered start-ups only (minimum 2 members)
*You will be added to the Shell.ai Hackathon startup edition 2025 page in the background within 7 days of filling the registration form. The timelines for the Special Start-up Edition will be available on the .
Shortlist and winners:
- Level 1 - Participants submit a pitch deck detailing their solution. Judges will select shortlisted start-ups, based on the evaluation of the pitching decks.
- Level 2 - Shortlisted start-ups will develop a prototype, based on their solution. The finalists (maximum three) will be announced based on the judging criteria.
- In Level 3 - The finalists will pitch their solutions live and the winners will be selected.
Prizes:
- The winning startup gets an opportunity to collaborate with Shell to develop a proof of concept, up to a value of US $50,000, for the winning startup’s submission.
All participants can register for the General Edition of the hackathon. Start-ups can register for General Edition. However, in this scenario, they will be eligible for General Edition awards only.
Participants interested in and eligible for the Special University can register for either General or Special University Edition, as they will be automatically eligible for the General Edition prizes, if they make it to the top three teams (General and Special University Edition combined) after Level 3.
Shell.ai Hackathon Timeline
| Date | Competition stage |
|---|---|
| 11th June | Hackathon opens for expression of interest to participate. |
| 4th July | Start of Level 1, Open for competition || Launch Webinar |
| 21st July | End of Level 1 (Start-up Edition) |
| 23rd July | End of Level 1 (General and University Edition), registrations close |
| 25th July | Level 1 shortlist announced, start of Level 2 |
| 18th August | End of Level 2 |
| 22nd August | Finalist announcement |
| 16th & 17th September | Live pitching session and winners announcement |
Hackathon Levels:
Level 1
General and Special University Edition
The participants are invited to develop mathematical models to optimise fleet decarbonisation strategies. 20 Participants with top scores in the private leaderboard (regardless of the edition) will be shortlisted for Level 2. More details on the challenge will be published in the problem statement.
Special Start-up Edition
Participants submit a pitch deck detailing their solution. Based on the judging criteria, pitch decks will be reviewed and a shortlist will be selected.
Level 2
All Editions
The shortlisted teams develop a prototype, based on their solution.
Level 3
All Editions
The finalists will pitch their solutions live and the winners will be selected.
Frequently Asked Questions
Our Partner for 2025
Winners
Shell.ai Hackathon for Sustainable and Affordable Energy 2025
Meet the winners of Shell.ai Hackathon for Sustainable and Affordable Energy 2025
Shello There Again: Winner General Edition
Team Members: Vibin M Vinod
Solution Approach: Team Shello There Again based out of Singapore developed a two-stage mixed integer linear programming (MILP) model for fleet optimization. The first stage optimized fleet size by assuming maximum yearly travel distance for vehicles. The second stage fine-tuned the distance distribution to minimally meet demand, significantly reducing costs. An intuitive web application was built with three key features: dataset management, cloud-based optimization leveraging AWS for scalability, and data analysis enhanced by an AI assistant for generating visualizations. This approach streamlined fleet management and empowered users with robust, scalable, and insightful decision-making tools.
Fleet-Opt: 1st Runner-up General Edition
Team Members: Hariharan Subramanian, Pranjal Mishra, Sachin, Gaurav Beswal
Solution Approach: Team Fleet-Opt based out of India utilized Mixed Integer Programming with Python and CPLEX, achieving solutions with a 0.5% optimality gap in under a minute. They developed a tuner page for fleet owners to adjust parameters and generate multiple models. The second phase involved a full-stack solution with a NextJS frontend, PRISMA and PostgreSQL, and an Express server hosted on Azure. The app supports multiple users running several models simultaneously using a queue-based framework and accounts for regional variations in maintenance centers, grid reliability, and charging station availability. Azure's deployment solutions enabled seamless multi-user access.
Barbarian King: 2nd Runner-up General Edition
Team Members: Rajshekhar Singhania, Alok Patel, Srikaran Reddy, Aditi Kambli
Solution Approach: Team Barbarian King based out of USA and India combined data analysis with advanced Operations Research techniques. They applied constraints during data preparation to simplify the problem and used symmetry-breaking techniques to accelerate model convergence. A dynamic vehicle recommendation layer was developed for flexibility across different datasets and computational environments. Their product provides detailed yearly fleet cost breakdowns based on uploaded data, making fleet planning straightforward and tailored to client needs.
Orwiz: Winner University Edition
Team Members: Naitik Kariwal, Rocky Hazowari
Solution Approach: Team Orwiz based out of Indian Institute of Management, Lucknow and International Institute of Information Technology, Hyderabad India developed an MILP model to optimize fleet decarbonization strategies, ensuring compliance with annual emission targets while meeting complex supply-chain demands. By exploiting problem structures, they reduced the optimization search space. A functional prototype was built, enabling clients to customize model parameters and generate future forecasts based on current data. A dashboard offers visualizations and actionable insights for fleet owners across different years, drivetrains, and fuels, addressing energy challenges and contributing to a sustainable future.
Analytic-BD: Runner-up University Edition
Team Members: Mohammad Nuwaisir Rahman, Miftahul Zannat Mifta
Solution Approach: Team Analytic-BD based out of BRAC University , Bangladesh formulated constraints and a cost function, utilizing Gurobi as the solver. They introduced an innovative approach to remove non-linearity, allowing for a fast initial feasible solution by substituting the decision variable regarding yearly distance traveled with the maximum yearly distance. This simplification accelerated the optimization process, making it easier to solve the problem and proceed with final optimization steps.
Nuralix.ai: Winner Startup Edition
Solution Approach: Team Nuralix.ai based out of USA developed Decarbo-Nix - An AI solution for fleet decarbonization, featuring a modular MILP engine that allows fleet managers to assemble custom optimization models. Powered by the DLT® framework, it integrates real-time asset monitoring, versatile data ingestion, and dynamic scenario management. Tailored dashboards and alerts provide instant visibility into fleet performance, emission reductions, and cost efficiencies. Scalable to fleets of any size, Decarbo-Nix drives carbon emissions down and profitability up, without sacrificing operational excellence.
Optexity: 1ˢᵗ Runner-up Startup Edition
Solution Approach: Team Optexity based out of India, optimized long-term decarbonization plans by approximating the problem as a linear programming model and removing quadratic terms. They improved solution speed and quality through hyperparameter tuning, problem-specific cutting planes, and iterative warm starts. Practical implementation focused on robustness through sensitivity analysis and quick replans for various scenarios. The final solution supports custom constraints, database integrations, and offers AI-driven analysis, visualization, and monitoring capabilities.
Ormae: 2â¿áµˆ Runner-up Startup Edition
Solution Approach: Team Ormae based out of India used a Stochastic Optimization modeling framework to address fuel cost variability. They leveraged the 'Branch and Bound and Cutting plane' algorithm in the Gurobi solver, a state-of-the-art method for solving large-scale complex optimization problems. This approach combines branch and bound with cutting planes to find optimal solutions to linear programming problems with integer constraints, ensuring efficient and effective fleet decarbonization.
Shello There Again: Winner General Edition
Team Members:
Vibin M Vinod
ShelloThereAgain! For fleet optimisation, I developed a two-stage mixed integer linear programming (MILP) model. In the first stage, I optimised the fleet size by assuming the vehicles travelled the maximum possible yearly distance. In the second stage, I fine-tuned the distribution of distance each vehicle should cover to minimally meet demand, leading to significant cost reductions.
I then built an intuitive web application offering three key features:
- Dataset Management: Users can easily view, edit, and upload their own datasets, providing flexibility in model inputs.
- Cloud-based Optimisation: By leveraging AWS architecture, the platform ensures scalability, whether for handling high user volumes or solving complex optimisation tasks.
- Data Analysis: Enhanced by an AI assistant that generates visualizations based on user input, making it easier to understand the proposed solutions.
In summary, this approach not only streamlined fleet management but also empowered users with robust, scalable, and insightful decision-making tools.
Fleet-Opt: 1ˢᵗ Runner-up General Edition
Team Members:
HARIHARAN SUBRAMANIAN
PRANJAL MISHRA
Sachin
Gaurav Beswal
For this challenge, we used Mixed Integer Programming with Python and CPLEX, and achieved solutions with a 0.5% optimality gap in under a minute. We developed a tuner page that allows fleet owners to adjust parameters, generate multiple models, and select optimal transition strategies.
For the second phase, we implemented a full-stack solution with a NextJS frontend, PRISMA and PostgreSQL, and an Express server, hosted on an Azure instance. This included a customized results page and a model comparison feature.
The app enables multiple users within an organization to run several models simultaneously using a queue-based framework. We also extended the model to a location-based framework, accounting for regional variations in maintenance centres, grid reliability, and charging station availability.
We thank Azure (Microsoft Inc.) for their seamless deployment solutions, enabling multi-user access to our app, and for waiving fees in support of innovation and scientific research.
Barbarian King:2â¿áµˆ Runner-up General Edition
Team members:
Rajshekhar Singhania
Alok Patel
Srikaran Reddy
Aditi Kambli
Our solution description:
We combined data analysis with advanced Operations Research techniques to simplify the complexity of the problem. Instead of embedding certain constraints directly into the model, we applied them during the data preparation stage, making the process more efficient. We also used symmetry-breaking techniques to accelerate model convergence. To ensure flexibility across different datasets and computational environments, we developed a dynamic vehicle recommendation layer. This not only improved performance but also allowed our model to adapt quickly, delivering near-optimal solutions in various scenarios, including the use of open-source solvers as an alternative to commercial ones. Our product provides clients with detailed yearly fleet cost breakdowns based on their uploaded data, making fleet planning straightforward and tailored to their needs.
Orwiz: Winner University Edition
Team Members:
Naitik Kariwal
Rocky Hazowari
We developed an MILP model to optimize fleet decarbonization strategies, ensuring compliance with annual emission targets while meeting complex supply-chain demands and constraints. By exploiting certain structures of the problem, we were able to reduce the optimization search space. We then built a functional prototype that enables clients to customize model parameters and provides real-time adaptability by allowing them to generate future forecasts based on current data. Additionally, we created a dashboard that offers visualizations and actionable insights for fleet owners across different years, drivetrains, and fuels. Our solution tackles pressing energy challenges and contributes to a more sustainable and affordable energy future.
Analytic-BD: Runner-up University Edition
Team Members:
Mohammad Nuwaisir Rahman
Miftahul Zannat Mifta
Nuwaisir about the solution approach:
To find the optimal solution of the fleet decarbonization challenge, we formulated the constraints, and a cost function, and then utilized Gurobi as our solver. Initially, one might pass the constraints and objective function directly into Gurobi. However, our observations indicated that this approach can be time consuming in reaching a good solution, primarily due to the presence of non-linear constraints. Eliminating this non-linearity could potentially accelerate the process of finding a feasible solution. Our method introduces an innovative approach to remove this non-linearity, allowing us to quickly obtain an initial feasible solution. We achieved a fast initial solution by substituting the decision variable regarding the yearly distance traveled by a vehicle with the maximum yearly distance it can travel, thereby simplifying the optimization task and making it easier to solve compared to the original problem. Then we proceeded with the final optimization steps starting from the initial solution.
Nuralix.ai: Winner Startup Edition
Decarbo-Nix by is a game-changing AI solution that puts fleet decarbonization on autopilot. Its modular MILP engine is like building with Lego blocks—fleet managers can assemble custom optimization models with just a click, effortlessly adapting to shifting sustainability goals. Powered by the cutting-edge and award winning DLT® framework, Decarbo-Nix seamlessly integrates real-time asset monitoring, versatile data ingestion, and dynamic scenario management, delivering unparalleled insights. With tailored dashboards and alerts, businesses gain instant visibility into fleet performance, emission reductions, and cost efficiencies—all in real-time. Scalable to fleets of any size, Decarbo-Nix is the ultimate tool to drive carbon emissions down and profitability up, without sacrificing operational excellence. Ready to transform your fleet’s sustainability journey? is the future you’ve been waiting for.
Optexity: 1ˢᵗ Runner-up Startup Edition
Optimizing the long-term decarbonization plan was an exciting challenge, and we were thrilled to top the global public leaderboard. In the first phase, we tackled the problem by approximating it as a linear programming model, removing quadratic terms. To further improve solution speed and quality, we used hyperparameter tuning, problem-specific cutting planes, and iterative warm starts. Later, we focused on practical implementation, ensuring robustness through sensitivity analysis and quick replans for various what-if scenarios. Our final solution supported custom constraints, database integrations and offered stakeholders AI-driven analysis, visualization, and monitoring capabilities.
Ormae:2â¿áµˆ Runner-up Startup Edition
We have used Stochastic Optimization modelling framework to address variability in fuel costs. We have leveraged ‘Branch and Bound and Cutting plane’ algorithm in Gurobi solver. This algorithm is a state-ofthe-art algorithm to solve large scale complex optimization problems. Branch and Cut is a method for solving mixed integer linear programs (MILPs), where some or all unknowns are integers, by combining branch and bound algorithm and cutting planes. It is used in combinatorial optimization to find optimal solutions to linear programming problems with integer constraints. It runs a branch and bound algorithm and tightens linear programming relaxations using cutting planes. It is a powerful technique for solving difficult MILPs that is used in many commercial optimization solvers like Gurobi, CPLEX, FICO Xpress.
