MSc Researcher · Ontario Tech University
AI engineer & graduate researcher working on trustworthy NLP for Requirements Engineering — LLM governance, knowledge discovery, and empirical software engineering.
MSc researcher working at the intersection of NLP and Requirements Engineering (RE). My thesis builds a four-study pipeline that takes raw stakeholder prose to structured, audit-grade knowledge artifacts — spanning LLM-based parsing with tag governance, transformer classification with local-context modeling, and multi-signal implicit keyword discovery grounded in an industrial FinTech/SaaS corpus of 110 requirements and 1,997 HLJ artifacts. Across three papers (CASCON 2025 published, RE'26 Main under review, RE'26 RE@Next! in revision), I argue that governance and distributional alignment — not raw model scale — are the primary levers for trustworthy RE automation.
Current Focus
Thesis in progress
MSc, Software Engineering · Ontario Tech University
Supervisor: Prof. Sanaa Alwidian
A four-study pipeline that automates requirements engineering (RE) using NLP and AI — from raw stakeholder prose to implicit domain knowledge discovery. The work spans structured JSON parsing, transformer-based classification with local context, and multi-signal keyword discovery, grounded in an industrial FinTech and SaaS corpus.
What I'm building
Multi-model pipeline (GPT-4.1, Claude Opus 4, Meta-70B) that converts raw requirements into confidence-scored HLJ artifacts. A 7-stage governance layer — harvest, filter, cluster, validate, whitelist, audit, drift — takes v2 precision to 0.95 and catches prompt-leak exploitation before it reaches downstream stages.
Empirical study across MPNet and DeBERTa-v3, frozen / LoRA / full fine-tuning, and context windows k=0–3. Structured divergence features move F1 by +16 points; distributional alignment beats scale (4K domain samples ≥ 15K mixed).
Five-phase engine: domain dictionary (13,725 entries), enriched co-occurrence graph (275K edges), three discovery signals — neighbor transfer, graph walks, UMAP+HDBSCAN gap detection — gated by a bounded LLM-as-judge that only tiebreaks the borderline band.
Every tag and every implicit keyword carries full provenance: discovery signal, per-signal scores, neighbor IDs, graph paths, cluster membership, SBERT similarity, dictionary match, and LLM judgment when invoked. Drift detection runs across pipeline versions.
The throughline across all three papers: governance and distributional alignment, not raw model scale, are the primary levers for trustworthy RE automation. LLMs are powerful but don't belong in every role — in this work, they parse and they tiebreak, but they never generate ground-truth knowledge without an audit trail.
Research Network
Explore the relationships between papers, research questions, datasets, and findings. Click a node for details, drag to rearrange, scroll to zoom.
Interactive Exploration
Adjust pipeline parameters and see how different configurations affect model performance. Results are pre-computed from actual experiment runs.
Adjust parameters to explore pre-computed results
Results
F1
84.7%
Precision
94.9%
Recall
79.9%
Confusion Matrix
Selected Work
Dvip Patel, Sanaa Alwidian · CASCON 2025, 2025
A modular multi-model LLM pipeline that converts raw stakeholder requirements into High-Level JSON (HLJ) artifacts with confidence-scored fields, paired with a versioned tag-governance system that catches prompt-leak exploitation, hallucinated tags, and low-agreement outputs before they reach downstream stages.
Dvip Patel, Amarachi Nwosu, Sanaa Alwidian · RE'26, RE@Next!, 2026
Explicit keyword extraction — even at audit-grade precision — hits a structural ceiling of roughly 1.5 canonical keywords per HLJ artifact, with Jaccard agreement below 0.11 across KeyBERT, RAKE, and YAKE. This paper presents a 5-phase implicit keyword discovery engine combining a 13,725-entry domain dictionary, a 275,164-edge enriched co-occurrence graph, UMAP+HDBSCAN clustering, and a bounded LLM-as-judge that tiebreaks only the borderline scoring band. Every implicit keyword traces back to its discovery signal(s) with full per-signal evidence.
Dvip Patel, Sanaa Alwidian · RE'26, Main Track, 2026
An empirical study of sentence-level requirement classification on 110 real-world FinTech and SaaS documents (~5,700 candidate sentences, balanced to 15K mixed / 4K domain-only). We compare all-mpnet-base-v2 (110M) and DeBERTa-v3-base (184M) across frozen / LoRA / full fine-tuning and context window sizes k=0–3, and show that structured local-context features — not raw concatenation — are the critical signal.
Where I've worked
Oshawa, ON · Supervisor: Prof. Sanaa Alwidian
Remote
Remote
India
Built & Shipped
Autonomous 10-step AI migration pipeline for Laravel codebases.
CLI-driven migration system that ingests legacy Laravel code and runs analysis → planning → transformation → validation → self-healing loops. Processes 10k+ LOC per run at under $80 cost, with 85–90% automated issue resolution.
Four-study pipeline for automated Requirements Engineering — structured parsing, context-aware classification, and implicit keyword discovery.
End-to-end research system behind my MSc thesis at Ontario Tech University. The pipeline converts raw stakeholder prose into audit-grade knowledge artifacts across four studies: 1. **Studies 1 & 2 (CASCON 2025)** — Multi-model LLM parsing (GPT-4.1, Claude Opus 4, Meta-70B) into High-Level JSON artifacts with versioned tag governance (v0 → v2; v2 precision 0.95 / F1 0.85). 2. **Study 3 (RE'26 Main, under review)** — Sentence-level requirement classification with structured local-context features (+16 F1 at k=1; best F1 0.894 on 4K domain-aligned samples). 3. **Study 4 (RE'26 RE@Next!, in revision)** — 5-phase implicit keyword discovery over a 275,164-edge enriched co-occurrence graph, UMAP+HDBSCAN cluster-gap detection, and a bounded LLM-as-judge tiebreaker. 4. **Cross-cutting infrastructure** — 1,997 HLJ artifacts, 13,725-entry domain dictionary, FAISS index over 768-dim SBERT vectors, full per-decision audit logging, and drift monitoring across pipeline runs.
The Toolkit
Technical competencies
Requirements Engineering, NLP for Software Engineering, LLM Governance, Knowledge Discovery, Empirical SE
Python, TypeScript / JavaScript, PHP (Laravel), SQL, Bash
PyTorch, HuggingFace Transformers, SBERT / MPNet, DeBERTa-v3, FAISS, LoRA Fine-tuning, LLM Applications, RAG Pipelines, UMAP + HDBSCAN
Distributed Pipelines, Pipeline Orchestration, Feedback Loops, Constraint-based Validation, Agentic Systems
Laravel, Flask, REST APIs, Microservices, RBAC, Docker, AWS, CI/CD, GitHub Actions
Education
Thesis: “Toward Automated Requirements Engineering: Empirical and Architectural Foundations for Structured Parsing and Knowledge Discovery”
Certifications
Let's talk
I'm open to research collaboration, AI engineering work, and good conversations about NLP for software engineering. The fastest way to reach me is email.