Hackathon statements

Build an autonomous agent that ingests an unfamiliar public-interest repository, localizes faults, tests multiple root-cause hypotheses, and produces cause-level patches. Validate every patch with synthesized property-based tests and mutation testing, and include a reviewer-ready explanation.

Expected outcome

AI Agents · Software Engineering · SDG 9, 16 Core challenges: Spectrum-based fault localization; multi-hypothesis reasoning; patch-overfitting detection; regression-test synthesis; compute-budgeted search. Success metrics: Verified bugs resolved; regression rate; mutation score; cost per successful repair; maintainer review time saved.

Build a recovery engine for damaged DVR/NVR drives and corrupted disk images that combines filesystem-aware recovery, carving, non-contiguous fragment reconstruction, and metadata rebuilding. Classify every output, attach confidence, preserve a hash-verified chain of custody, and generate a court-ready report.

Expected outcome

Digital Forensics · Systems Programming · SDG 16 Core challenges: Fragment-gap reconstruction; proprietary DVR containers; timestamp and metadata forensics; read-only evidence handling; provenance logging. Success metrics: Recovery precision; file completeness; false-positive rate; metadata accuracy; evidence-integrity verification pass rate.

Build a platform that learns a live service-dependency graph, separates transient noise from incidents, localizes root causes from logs, traces, and metrics, and executes recovery under blast-radius limits, canary remediation, and automatic rollback.

Expected outcome

Distributed Systems · AIOps · SDG 9, 16 Core challenges: Causal inference over dependency graphs; change-point detection; safe-action policy engines; cascade prediction; rollback. Success metrics: Detection accuracy; mean time to identify and recover; false-alert rate; remediation-induced regressions (target: zero).

Build an authorized, hard-scoped security agent that maps an attack surface and tests authentication, authorization, business logic, and multi-endpoint attack chains. Every confirmed finding needs a reproducible non-destructive proof, sensitivity-aware risk rating, fix recommendation, and automated retest.

Expected outcome

Application Security · AI Agents · SDG 16, 10 Core challenges: Stateful multi-role exploration; authorization-differential testing; attack-chain planning; non-destructive proof generation; fix verification. Success metrics: Confirmed vulnerabilities; severity-weighted detection; reproducibility; false-positive rate; retest accuracy.

Design a fully offline system for clinics, community health workers, or village schools that later merges concurrent edits over weak links without silently losing a valid record. Unsafe automatic merges must surface an explicit conflict.

Expected outcome

Distributed Systems · Database Engineering · SDG 3, 4, 10 Core challenges: CRDTs and hybrid logical clocks; domain-aware merge semantics; resumable delta sync; idempotent replay; device-loss recovery; encryption at rest. Success metrics: Data-loss rate (target: zero); sync latency; bytes per sync; conflict-resolution correctness; interrupted-sync recovery.

Generate SBOMs and real dependency graphs, then use call-graph reachability to determine whether vulnerable functions are exploitable. Detect malicious or typosquatted packages, verify build provenance, and produce an evidence-backed remediation plan with machine-readable VEX.

Expected outcome

DevSecOps · Static Analysis · SDG 9, 16 Core challenges: SPDX/CycloneDX SBOMs; static reachability; malicious-package detection; provenance validation; risk prioritization. Success metrics: Detection precision; noise eliminated; actionable-finding rate; dependency-impact accuracy.

Build a verifiable-credential system that proves only a needed claim through selective disclosure or zero-knowledge predicates. Verification must work offline on low-end phones, support privacy-preserving revocation, and resist cross-verifier tracking.

Expected outcome

Applied Cryptography · Privacy Engineering · SDG 16.9, 1, 10 Core challenges: BBS+/SD-JWT selective disclosure; ZK predicates; unlinkability; privacy-preserving revocation; issuer trust; threat modelling. Success metrics: Verification correctness; forgery resistance; unlinkability guarantees; offline verification latency.

Choose an SDG-anchored crop, health, flood, or landslide inference task and run it fully offline on low-cost battery-powered hardware. Optimize the full pipeline under fixed memory, latency, and energy limits while maintaining cross-region and cross-device robustness.

Expected outcome

Machine Learning · Edge Computing · SDG 2, 3, 13 Core challenges: Quantization-aware training; pruning and distillation; hardware profiling; energy measurement; calibration; subgroup robustness. Success metrics: Accuracy; latency; peak memory; accuracy per joule; calibration error; worst-subgroup performance; deployment cost.

Build an investigator that reconstructs a cross-service failure timeline despite incomplete logs and clock skew, ranks root-cause hypotheses, and tests them counterfactually. Reports must separate facts from inferences and link claims to evidence with calibrated confidence.

Expected outcome

Observability · AI Agents · SDG 9, 16 Core challenges: Cross-service correlation; clock-skew correction; causal analysis; missing-data reasoning; evidence provenance; confidence calibration. Success metrics: Root-cause and timeline accuracy; investigation time; evidence quality; confidence calibration.

Build a controller that forecasts demand, occupancy, and renewable generation, then optimizes HVAC, storage, and loads for cost and grid carbon intensity while protecting critical loads, comfort, battery life, and outage resilience.

Expected outcome

Optimization · IoT · Energy Systems · SDG 7, 11, 13 Core challenges: Probabilistic forecasting; model-predictive or safe-RL control; carbon-aware scheduling; islanding; anomaly detection; digital twins. Success metrics: CO2e and cost reduction; peak-demand reduction; forecast error; comfort and critical-load constraint violations.

Using synthetic or properly authorized data, build a streaming engine that detects suspicious activity and mule rings through graph analysis, adapts to drift under severe class imbalance, and produces explainable low-latency alerts without unfairly blocking vulnerable user segments.

Expected outcome

Machine Learning · FinTech · SDG 16.4, 1, 8, 10 Core challenges: Streaming graph analytics; drift adaptation; imbalance handling; explainability; fairness auditing; low-latency inference. Success metrics: Precision-recall; false-positive rate; latency; drift robustness; segment-wise false-decline gap.

Build a dynamic scheduler that profiles workloads and places jobs across shared CPU, GPU, and NPU resources while respecting SLOs, memory limits, tenant fairness, contention, and schedule stability. Shift flexible jobs toward low-carbon and low-energy windows.

Expected outcome

Operating Systems · Systems Engineering · SDG 7, 9, 13 Core challenges: Workload profiling; performance and energy prediction; carbon-aware placement; contention handling; multi-tenant fairness. Success metrics: Throughput; tail latency; utilization; energy and CO2e per job; fairness index; scheduling stability.