experience

2025 — ai researcher, algoverse ai research

> conducted ai research in a 12-week program with mentorship from phds at meta, openai, and princeton.

> enhanced large language models (qwen3-235b-a22b, gpt-oss) by implementing a self-contrastive mixture-of-experts (moe) architecture using pytorch and jupyter notebooks.

> co-authored a research paper under review, introducing novel contrastive decoding methods achieving a 2% efficiency gain in moe models across benchmarks.

2025 — genai intern, government of canada – immigration board

> developed and tested an azure workflow to transcribe, classify, and organize refugee basis of claim documents with 93% accuracy, reducing manual processing time.

> engineered an azure-based chatbot leveraging cosmos db (nosql), nlp, and embedding pipelines, reducing legal document ingestion time to under 10 minutes.

> delivered a conference presentation translating ai workflows into actionable insights for stakeholders.

links