The Sequential Matching Problem

The Sequential Matching Problem: Romeo & Juliet × IIT Madras An Applied Machine Learning Programme & Hackathon How should a system decide who should be introduced to whom when it does not know everything about the people involved, feedback arrives over time, and each decision changes the options and information available later? The Sequential Matching Problem explores this question through an applied machine learning programme focused on matching, recommendation systems, optimisation and sequential decision-making. The programme is designed to go beyond a conventional hackathon. Participants will first explore the challenge through workshops and research before moving into implementation, experimentation and mentorship. The Challenge: The challenge explores how intelligent systems can make matching decisions when information is incomplete, feedback changes over time, and each decision can affect what happens next. Participants will explore questions such as: How should a system make decisions with incomplete information? How should it respond when feedback arrives over time? How should uncertainty be handled? What information is worth acquiring? How should different possible outcomes be compared? How can a system make decisions that remain effective as conditions change? The programme brings together ideas from machine learning, recommendation systems, optimisation and sequential decision-making. Participants will work with a controlled environment using synthetic data, allowing them to experiment with these ideas without using real-world personal data. How The Programme Works: The programme is structured across three stages. 1. Workshops & Research: 5 to 11 October 2026 The first stage focuses on learning, research and technical reasoning. Participants will attend technical sessions, explore the challenge and work with their teams to develop their approach. Teams will investigate: Problem interpretation Technical assumptions Modelling approaches Decision strategies Uncertainty Experimental design Potential limitations and failure cases Teams will then submit their research and proposed approach before moving into the build phase. 2. Build & Mentorship: 12 to 18 October 2026 Selected teams will move into the implementation phase. Teams will gain access to the development environment and turn their ideas into working systems. During this phase, teams will: Implement their approach against the simulator Compare their approach with supplied baselines Run experiments and ablations Investigate performance under different conditions Study the effects of incomplete, noisy or delayed information Take part in technical mentorship and office hours Prepare a reproducible final submission 3. Final Evaluation: Following the build phase, selected teams will enter the final evaluation process. Systems will be evaluated in controlled environments, including private environments that are not identical to the public development environment. This is designed to assess how well an approach generalises beyond the development setting. Finalists will also present their work and answer technical questions about their approach, experiments and findings. The Development Environment: The competition will use a Python-based sequential simulator with documentation and baseline examples. The simulator provides a controlled environment where teams can make decisions, observe outcomes and study how their systems behave over time. The underlying outcome-generating process will remain hidden from participants. Data & Privacy: The competition uses synthetic data throughout. Participants will not be asked to provide dating, relationship or compatibility information. No real Romeo & Juliet member data will be used. Participants retain ownership of the code, models, reports and other work they submit. Participation does not transfer ownership of a submission to Romeo & Juliet. Romeo & Juliet retains ownership of its own platform, technology, materials and work developed by or for the company before, during or after the programme. Participation, or a similarity between a submission and Romeo & Juliet’s work, does not give a participant ownership of Romeo & Juliet’s independently developed work. If Romeo & Juliet wishes to use a participant’s submitted work in its product, the parties will discuss that separately and agree the terms in writing. Who Can Participate?: The programme is open to students from IITs across India. Eligibility: Any year of study Any academic discipline or degree programme Teams of 1 to 4 students You do not need to be enrolled in a formal machine learning course. Participants should be comfortable with at least one quantitative or technical area such as coding, mathematics or data analysis. Recommended Preparation: The following are recommended, but not mandatory: Python Basic probability and statistics Introductory machine learning Students from computer science, data science, mathematics, physics, engineering, economics and related disciplines are encouraged to participate. Programme Timeline: Phase Dates Registration & Team Formation 25 Sep to 2 Oct 2026 PS Release 5 Oct 2026 Workshops & Research 5 to 11 Oct 2026 Build & Mentorship 12 to 18 Oct 2026 Final Evaluation After 18 Oct 2026 Final Presentations & Judging Date to be announced Format: Fully remote Team Size: 1 to 4 students Data: Synthetic throughout What You Will Submit: Round 1: Research & Approach Teams will develop and submit their proposed approach, covering areas such as: Problem interpretation Key hypotheses Proposed modelling approach Decision or allocation strategy Treatment of uncertainty Learning strategy Planned experiments Relevant baselines Expected failure modes Build Stage: Selected teams will submit: Working implementation Source code or repository Reproducible environment Experimental and ablation results Comparison against supplied baselines Short technical report Final Submission: Finalists will be expected to provide: Final reproducible system Technical report and experimental analysis Private evaluation execution Presentation or demonstration Technical Q&A GitHub repository URL Workshops & Technical Learning: The programme will explore topics across applied machine learning and technical decision-making. Reciprocal Matching & Recommendation Systems: Two-sided recommendation, directional outcomes and the difference between ranking pairs and allocating across a pool. Learning From Noisy & Incomplete Data: Missing information, uncertainty, sparse observations and cold-start scenarios. Optimisation Under Constraints: Global matching, constrained optimisation and computational trade-offs. Exploration, Exploitation & Online Learning: Selective feedback, information value and sequential learning. Building Robust ML Systems From Messy Data: Experimental design, validation, reproducibility and failure analysis. From Technical Model To Real Product: Connecting technical performance with practical constraints and user value. How Will Teams Be Evaluated? Evaluation will consider the quality and performance of the systems developed throughout the programme. Areas may include: Technical performance Matching and allocation quality Handling of uncertainty and incomplete information Robustness Experimental quality Research reasoning Reproducibility Detailed evaluation criteria will be provided as part of the competition documentation. Beyond The Hackathon: The programme also provides an opportunity to connect with the technical team at Romeo & Juliet. Selected participants may be considered for paid internship opportunities with Romeo & Juliet. Internship consideration may take into account technical work, independent problem-solving, experimentation, scientific and engineering judgement, ability to explain technical trade-offs and subsequent technical discussions or interviews. Exact internship terms will be communicated separately once finalised. Registration: There Are Two Steps To Complete Your Registration Step 1: Register On Unstop Create or join your team on Unstop. After forming or joining your team on Unstop, proceed to the official event website and complete the registration process. You will provide the required participant and team information and complete a short interaction with Juliet, Romeo & Juliet's AI onboarding agent. The interaction takes approximately five minutes and consists of six questions. There are no right or wrong answers.