I build practical Python tools for research, education and career decision-making. My background is in materials chemistry and science teaching; my current focus is turning that domain knowledge into useful, inspectable AI applications.
UK-based · Open to AI engineering, RAG, agentic AI and Python roles across the UK, including remote and hybrid.
| Project | Problem it addresses | Technical evidence |
|---|---|---|
| Job Evidence Mapper | Connects job requirements to truthful candidate evidence and highlights gaps for human review. | Python, Streamlit, sentence-transformers, semantic similarity matching, CSV/Markdown exports and regression tests. |
| IT Support Operations Agent | Prepares source-linked replies to fictional IT support tickets for human review. | LangGraph interrupt/resume, scoped TF-IDF retrieval, FastAPI and Streamlit, 13 tests and authored workflow evaluations. Try the fictional demo. |
| STEMSpark Event Planner | Plans school STEM events with budget, age-suitability and accessibility checks. | Pydantic models, deterministic validation tools, bounded plan revision, audit traces and tests. |
| RareEarthRAG | Helps researchers find relevant passages in rare-earth materials papers. | PDF text extraction, TF-IDF retrieval, cosine similarity and source-linked results. Currently a retrieval baseline, not a complete generative RAG system. |
| SynthAudit | Checks whether synthesis methods report key details for expert review. | Modular Python rules, evidence snippets, guardrails, tests and GitHub Actions. Deterministic extraction, not an LLM. |
| StemSplit AI | Separates uploaded music into six audio stems using a pretrained model. | FastAPI, Demucs integration, file-processing endpoints and Docker packaging. Local MVP with documented production limitations. |
The linked repositories include project blueprints, architecture notes, test reports and known limitations. Materials Paper Summariser and PrivatePDF to Word provide additional document-processing examples.
- Research depth: PhD in Materials Chemistry, IIT Madras; postdoctoral research at the University of Southampton. My research focused on rare-earth molybdoantimonites, structural characterisation and luminescence.
- Clear communication: science teaching experience and a PGCE with QTS, helping me translate technical ideas into understandable workflows.
- Practical engineering: Python applications with explicit validation, inspectable outputs and documented limitations.
- Human oversight: tools designed to support judgement, surface uncertainty and avoid presenting heuristic scores as facts.
I am studying the Professional Certificate in Agentic AI from IITM Pravartak, delivered with Emeritus.
My learning focus includes retrieval-augmented generation, embeddings, agent workflows, evaluation and deployment. The projects above distinguish what is implemented today from planned extensions.
- SciFigAudit — provides explainable technical and accessibility signals for scientific figures before human review.
My approach: start with a real problem, build a transparent baseline, test its behaviour and document what it cannot do.