Darshini Balamurali

Software Engineering graduate with a strong foundation in Artificial Intelligence and machine learning

Talent

WinnipegMember since June 13, 2026

PythonJavaReact.jsNode.jsFastAPIFlaskMulti-Agent SystemsGATCodeBERTDiffusion ModelsHopfield NetworksOpenAI CLIPSAM2SpaCyLightGBMXGBoostSARIMAXSHAPNumPyPandasNeo4jMongoDBMicrosoft AzureCode ReviewsPyTorchEnergy-Based ModelsIBM Watson NLUStatsmodelsTableauMatplotlib

Hire this person

Send a job offer directly to this candidate

PythonJavaReact.jsNode.jsFastAPI

About

Software Engineering graduate with a strong foundation in Artificial Intelligence and machine learning, with practical experience working across full stack, designing and deploying backend systems and APIs, managing relational and graph databases, building cloud-hosted applications, and developing ML systems ranging from classical ensemble models to graph neural networks, large language model pipelines, and generative AI research. Having contributed equally to software engineering codebases and designed and evaluated machine learning systems end to end, I have built solid experience with modern development practices from sprint planning and code reviews through to production deployment and testing.

Experience

AI/ML Research Intern - Jaguar Land Rover - University of Windsor, Windsor, Canada

(2025-09 - 2025-12)

Engineered a two-phase static analysis framework to flag performance-bug-inducing commits across large-scale open-source repositories (Jenkins, PyTorch), processing a combined dataset of 220,759+ function-level instances.
Developed a commit-level ML pipeline using XGBoost and LightGBM with TF-IDF feature extraction and SZZ-algorithm-based bug labeling, attaining ROC-AUC of 0.91 and PR-AUC of 0.92 on a 4,100-commit balanced dataset.
Constructed multi-level graph representations (commit DAGs, file-version graphs, function-evolution graphs) across 99,400+ nodes, capturing code evolution patterns invisible to commit-level metadata.
Implemented AST parsing pipelines using Tree-sitter for Java and Python, extracting 220,000+ function snapshots from 3,325 source files.
Developed a GAT + MLP ensemble embedding CodeBERT with AST structural features, scoring PR-AUC of 0.885 and cutting false positives versus the standalone GAT, deployed via Flask with SHAP-based explainability.

Industry Project Intern - Meallens - University of Windsor, Windsor, Canada

(2025-05 - 2025-08)

Engineered a full-stack web platform combining SAM2 and OpenAI CLIP for semantic search and intelligent content understanding across a structured dataset.
Automated team notifications via Slack API integration and powering real-time operational alerts across a distributed team.
Designed and deployed interactive dashboards with data visualizations, allowing stakeholders to surface key performance metrics.
Shipped iterative feature releases on schedule across multiple Agile sprint cycles, contributing to planning, code reviews, and retrospectives throughout the engagement.