Prospective PhD researcher in trustworthy AI, medical machine learning, and multimodal evaluation
I am an MCA graduate from the National Institute of Technology Kurukshetra (CGPA 8.6/10) and a Technical Trainer at Chandigarh University. My current work connects careful evaluation with deployable AI systems:
- explainable learning for imbalanced gastrointestinal-disease classification;
- cluster-aware evaluation of hosted vision-language models for tactile-property classification;
- reproducible software for testing robustness, calibration, retrieval quality, and failure modes.
I have taught data structures, cloud computing, and full-stack engineering to more than 600 students. That experience shapes how I document experiments: assumptions should be visible, baselines should be reproducible, and limitations should be stated plainly.
| Work | Research question | Status |
|---|---|---|
| Explainable ML for gastrointestinal-disease classification | How should Borderline-SMOTE B2, Random Forest, and explanation-aware analysis be evaluated when clinically important classes are underrepresented? | Manuscript in preparation |
| Cluster-aware evaluation of hosted vision-language models | How does tactile-property classification performance change when evaluation respects cluster structure rather than treating every sample as independent? | Manuscript in preparation |
| Repository | What it demonstrates |
|---|---|
| MedVision Reliability Lab | A medical-image reliability workbench with imbalance comparisons, validation-only calibration, external-centre testing, appearance-cluster failure slices, and faithful linear explanations. |
| SciEvidence RAG Audit | An evidence-retrieval and scientific-claim audit pipeline with a measured SciFact baseline, source-disjoint calibration, abstention analysis, and inspectable citations. |
Each repository includes tests, continuous integration, dataset citations, limitations, and a project-defense guide. Reported numbers are generated by the checked-in evaluation pipeline; no benchmark result is claimed without a reproducible artifact.
- NIT Hostel Management System: production-oriented student and administrator workflows, containerized deployment, and automated delivery.
- DigiMine: an education platform with teacher portals and cloud deployment.
Research: class-imbalance handling, cluster-aware evaluation, calibration, model comparison, error analysis, and reproducible experimentation
AI systems: Python, scikit-learn, computer-vision feature pipelines, sparse retrieval, hosted LLM integration, and vector search
Engineering: Java, C++, Go, JavaScript/TypeScript, React, React Native, PostgreSQL, MongoDB, Qdrant, Docker, AWS, GitHub Actions, testing, and CI/CD
I am preparing for funded PhD work in Europe, particularly projects involving AI for health, trustworthy and explainable ML, multimodal or tactile perception, human-AI collaboration, and reliable AI software.