Hackforge

Ignite Your Potential: Inside HackForge @ F.E.T.S.U. presents Srijan'26   Welcome to HackForge @ F.E.T.S.U. presents Srijan'26 Forget standard coding competitions. HackForge is a crucible where raw ideas are hammered into groundbreaking tech. Hosted offline by CodeClub JUSL as the flagship hackathon of Srijan'26, this is the ultimate arena for creators, problem-solvers, and visionaries to build without limits. Why step into the Forge? Build with Impact: Whether you're spinning up a dynamic frontend, training an intelligent ML model, or designing a flawless user experience, this is your chance to tackle real-world problems and develop solutions that actually matter. The Ultimate Dev Ecosystem: Innovation doesn't happen in a vacuum. Connect with the sharpest minds across diverse academic backgrounds, find future collaborators, and experience the unmatched energy of a live, in-person hackathon. Push Your Stack to the Limit: HackForge is the perfect sandbox for everyone. Whether you are a seasoned developer deploying complex architectures or an aspiring tech enthusiast looking to test your skills under pressure, this is your stage to level up. Grab your team, fire up your IDEs, and bring your best ideas to the table. The future of technology is waiting to be built—let’s forge it together.   Guidelines and Rules Eligibility The hackathon is open to all students currently enrolled in a college or university. Team Formation Each team must consist of 3–4 members. Teams may include students from different institutions and academic batches. Teams not meeting the required size will be disqualified. Registration All team members must register individually before the submission deadline. Failure to complete registration will result in team disqualification. Competition Rounds The hackathon will consist of two rounds. Idea Submission Round Teams must submit a PDF of their Presentation (maximum 10 slides). The presentation must be based on the provided themes. The PPT must follow the PPT format provided. Please refer to this link to access the ppt format -https://www.canva.com/design/DAHCWpP_POI/tsSh32fiAMzsO89GIrL1lg/edit?utm_content=DAHCWpP_POI&utm_campaign=designshare&utm_medium=link2&utm_source=sharebutton Candidates can do the ppt in PowerPoint (preferred or in Canva by making a copy of the format given. Submissions must be made through the designated online portal before the deadline. Coding + Pitching Round Shortlisted teams will participate in the on-site round. Venue: SMCC Building, Jadavpur University Salt Lake Campus. The round will start at 10:00 AM sharp. All team members must be present on time. Late arrival will result in disqualification. Development Rules Participants may use any technology stack of their choice. All GitHub commits must be made only after the start of the on-site round. Any commits made before the start time will result in immediate disqualification. Teams must regularly push their progress to the designated repository. Code will be evaluated at multiple timestamps. Academic Integrity All submissions will undergo a plagiarism check. Any instance of code plagiarism or idea replication will result in disqualification. Facilities Food and beverages will be provided to all team members during the on-site round. Awards The best-performing teams will receive cash prizes and recognition certificates. All teams will receive a participation certificate for all members. Problem Statements: Decentralized Disaster Relief Governance & Fund Accountability Framework Eastern India, particularly Assam and West Bengal, faces recurring hydrological disasters that trigger large-scale public and institutional funding flows. Despite increasing digital payment penetration, the lifecycle of relief capital—from donation to beneficiary disbursement—remains opaque, fragmented across NGOs, and prone to administrative inefficiencies and trust deficits. Current Web2-based dashboards lack cryptographic guarantees, auditability, and participatory governance mechanisms. Design and implement a decentralized disaster-relief governance protocol that integrates programmable fund custody, milestone-based capital release, and community-verifiable proof-of-utilization. The system should enable cryptographically enforced allocation logic via smart contracts, decentralized storage of expenditure artifacts (geo-tagged images, invoices, logistics proofs), and DAO-based voting for tranche approvals. Consider privacy-preserving donation flows (optional anonymity via zero-knowledge proofs) and interoperability with digital public infrastructure. Teams should architect a scalable Layer-2 deployment strategy, incorporate account abstraction for frictionless onboarding, and demonstrate transparent capital flow visualization. The solution must address governance attacks, oracle manipulation risks, and Sybil resistance in participatory decision-making. Self-Sovereign Digital Identity & Verifiable Skill Credentials for Migrant Workforce Mobility Seasonal and long-term migrant workers from Bihar and Jharkhand frequently encounter systemic exclusion due to fragmented documentation, unverifiable skill claims, and lack of portable benefit eligibility records across state boundaries. Centralized identity repositories introduce surveillance risks, data misuse vulnerabilities, and bureaucratic latency. Develop a self-sovereign identity (SSI) ecosystem based on W3C Decentralized Identifiers and Verifiable Credentials, enabling workers to cryptographically own, manage, and selectively disclose employment records, vocational certifications, and welfare entitlements. The architecture must support zero-knowledge-based selective disclosure, revocation registries, and credential verification without persistent centralized storage of personal data. The system should model real-world issuer-verifier-holder trust relationships (e.g., training institutes, contractors, state bodies), incorporate revocation logic, and demonstrate interoperability with wallet-based mobile interfaces. Threat modeling against identity theft, replay attacks, and malicious issuers should be explicitly addressed. Multi-Stakeholder Blockchain Protocol for Agricultural Commodity Traceability in Eastern India Eastern India’s agrarian economy, especially in West Bengal and Odisha, is deeply dependent on rice production and multi-tier distribution networks involving smallholder farmers, aggregators, wholesalers, and retailers. Lack of provenance verification enables counterfeit labeling, adulteration, and price manipulation, while consumers remain disconnected from origin-level transparency. Architect a permissioned or hybrid blockchain-based agricultural traceability network that models commodity lifecycle states from cultivation to retail. The protocol should support batch tokenization of produce lots, cryptographically signed transfer-of-custody events, and immutable timestamped logs accessible via consumer-facing verification portals. QR-encoded state transitions should map to on-chain transaction hashes. The design must address scalability (high transaction frequency during harvest seasons), off-chain data integrity (IoT or manual input validation), consensus mechanism trade-offs, and role-based access control. Teams should demonstrate tamper-resistance, audit trace reconstruction, and resilience against collusion between supply chain actors. Climate-Aware Probabilistic Crop Yield Modeling Using Multi-Source Data Fusion Agricultural productivity across Eastern India is increasingly volatile due to non-linear climatic shifts, irregular monsoons, and soil degradation patterns. Small and marginal farmers lack predictive tools that integrate heterogeneous environmental variables into actionable forecasts. Design a multi-modal machine learning pipeline capable of probabilistic crop yield forecasting using structured (historical rainfall, soil pH, fertilizer usage) and semi-structured (satellite-derived vegetation indices, seasonal climate anomalies) datasets. The system must incorporate uncertainty quantification (confidence intervals or Bayesian modeling), risk stratification (low/medium/high yield), and interpretability mechanisms to ensure farmer trust. Participants should evaluate ensemble-based models, temporal forecasting architectures, and automated hyperparameter optimization strategies. Emphasis must be placed on handling missing rural datasets, mitigating bias from skewed historical records, and generating explainable agronomic recommendations rather than black-box predictions. Deployment should demonstrate real-time inference capability with resource-efficient serving mechanisms suitable for low-connectivity rural environments. Spatiotemporal Demand Forecasting & Food Waste Optimization in Urban Institutional Kitchens Urban institutions in Kolkata experience significant post-consumption food waste due to inaccurate attendance estimation, seasonal variability, and menu-dependent demand fluctuations. Traditional heuristic planning fails to capture latent temporal patterns and behavioral signals. Construct a time-series and feature-engineered ML framework for multi-horizon food demand forecasting. The system must integrate calendar effects, academic schedules, historical consumption patterns, weather variability, and menu attributes into a structured forecasting model. Explore advanced architectures such as Temporal Fusion Transformers or hybrid gradient-boosting ensembles with lag feature engineering. Beyond prediction, teams should implement an optimization layer that minimizes expected waste under uncertainty constraints. Include performance evaluation via RMSE/MAE and simulate cost-saving projections under various adoption scenarios. The system should demonstrate real-time retraining or adaptive learning capabilities. Deep Learning-Based Multi-Class Fish Freshness & Quality Assessment Using Computer Vision Fish markets across West Bengal and Assam rely heavily on subjective visual inspection to determine freshness, leading to inconsistent quality grading and potential public health risks. Manual evaluation cannot systematically account for subtle visual biomarkers such as eye clarity degradation, gill discoloration, and surface texture changes. Develop a deep convolutional neural network (CNN) or vision transformer-based classification system capable of multi-class freshness grading under real-world lighting variability. The model should incorporate data augmentation strategies, domain adaptation techniques for noisy market environments, and confidence-calibrated outputs. Participants must justify architecture selection (e.g., ConvNeXt, EfficientNetV2, lightweight ViTs), perform transfer learning from pretrained weights, and evaluate performance using precision-recall metrics and confusion matrices. Edge deployment constraints (mobile inference latency, quantization, ONNX export) should be considered. Advanced solutions may incorporate explainable AI methods (Grad-CAM, attention heatmaps) to highlight image regions influencing freshness classification, enhancing trust and usability among vendors and consumers. Try to accommodate each problem statement in one page Problem statement 7: If you have any idea that you want to implement as a problem statement that can solve the problem of poverty, hunger of the people, then please first send the mail along with your tech stack to Code Club JUSL, if the statement is well-advised, then you can work on that after approval. Mail id - codeclubjusl@gmail.com Please make your ppt in the preferred format and submit your idea!! Point of contacts: Aritra Mondal +91 7365911452, Dipan Mondal +91 8250821406, Vivek Haldar +91 98754 95117