Mangalore Refinery and Petrochemicals Limited (MRPL)
Browse and strategize across all 3 official problem statements submitted by Mangalore Refinery and Petrochemicals Limited (MRPL) for Smart India Hackathon 2026.
Problem Statements (3)
Match My TeamSovereign On-Premise Agentic AI Workbench using Open-Weight Multimodal LLMs for Confidential Industrial Work
Refineries (MRPL), defense PSUs, and sovereign government agencies handle highly sensitive knowledge work (refinery shutdown planning, crude blend optimizations, board memos, engineering calculations) that cannot be sent to foreign commercial cloud LLMs (OpenAI/Anthropic) due to national security. Build a Sovereign On-Premise Agentic AI Workbench for MRPL using open-weight multimodal foundation models (Llama-3, Qwen-VL, Mistral) running 100% air-gapped on local enterprise GPUs with multi-agent orchestration for engineering document synthesis.
Passive Colorimetric H2S Exposure-Dosimeter Wristband with AI-Based Quantitative Reading
Refinery workers at MRPL face chronic low-level Hydrogen Sulfide (H2S) toxic gas exposure (1-10 ppm) that causes neurological damage over time, which standard electronic gas detectors miss because they only sound alarms at acute emergency thresholds (>10 ppm) and require batteries. Build a Passive Colorimetric H2S Exposure-Dosimeter Wristband and AI-Based Quantitative Smartphone Reader that continuously accumulates chemical exposure on a metal-salt colorimetric film and uses smartphone camera computer vision with color calibration to calculate exact cumulative time-weighted average (TWA) dosage.
Indigenous GPU-Accelerated Optimization Solver (Sovereign Alternative to Express / CEPLEX)
Indian petroleum refineries (MRPL), petrochemical complexes, and power grids depend entirely on expensive foreign commercial mathematical optimization solvers (Gurobi, IBM CPLEX, FICO Xpress costing millions in annual licensing) for critical refinery LP scheduling and crude blending. Build an Indigenous GPU-Accelerated Mathematical Optimization Solver (Sovereign alternative to Gurobi/CPLEX) in C++/CUDA and Rust that solves large-scale Linear Programming (LP), Mixed-Integer Programming (MIP), and Quadratic Programming (QP) problems utilizing parallel Interior Point Methods (IPM) and GPU Branch-and-Bound.