AI-powered diagnostic system for analysing why applications fail and generating structured improvement feedback with predictive success estimation.
Failure Autopsy Engine (FAE) is an AI-driven evaluation support system designed to identify the root causes behind rejected applications and provide structured guidance for improvement.
Traditional evaluation systems only deliver binary outcomes — accepted or rejected — without explaining failure patterns. FAE introduces an analytical framework that diagnoses decision factors, compares submissions against successful benchmarks, and generates actionable feedback to improve future outcomes.
The system is designed for scalable evaluation environments such as:
grant applications
academic admissions
hiring processes
funding proposals
competition submissions
FAE transforms evaluation from outcome reporting to decision intelligence.
Most application review systems lack transparency and structured feedback. Applicants receive rejection decisions without understanding:
which evaluation criteria were not met
how their submission differs from successful cases
what specific improvements are required
This creates inefficiency, frustration, and repeated failure cycles.
FAE addresses this by analysing submission content, comparing it with successful benchmarks, and identifying improvement pathways using AI-assisted reasoning.
Root cause analysis of rejected applications
Semantic comparison with successful submissions
AI-generated structured feedback
Severity scoring of weaknesses
Predictive success estimation
Scalable evaluation architecture
The system is designed as a modular analytical pipeline:
- Document ingestion and preprocessing
- Semantic embedding generation
- Vector similarity search against benchmark database (FAISS)
- Pattern and gap identification
- AI-driven feedback generation using large language models
- Predictive scoring and recommendation output
This pipeline enables consistent, explainable evaluation support.
- Python
- Semantic embeddings
- FAISS vector similarity search
- Large Language Models (LLM-based feedback generation)
- Scalable cloud architecture design
Developed as part of the AWS AI for Bharat Hackathon (2026), focusing on building AI systems that support decision-making in real-world evaluation environments.
- Education and admissions evaluation
- Recruitment and talent screening
- Grant and funding review
- Policy and programme assessment
- Competitive selection processes
- Real dataset integration
- Model performance evaluation metrics
- Interactive web interface
- Human-in-the-loop evaluation controls
- Deployment-ready cloud infrastructure
Mohammed Imad Thotan
Mohammed Saif Sharif
Business Analytics | AI System Design | Decision Intelligence
Intellectual Property Notice This project concept, architecture, and implementation are original work. Unauthorized reproduction, distribution, or commercial use is prohibited.