Mohamed Nayeem

Fine-tuned and deployed LLMs (DeepSeek, Mistral, Qwen, Falcon, BERT variants) to improve
product attribute extraction and compatibility mapping, achieving ~15% higher classification
precision in retail catalog pipelines.
Developed an internal Streamlit tool that used RAG + Redis to extract product brand and
model from bulk product text (100+ items), reducing manual labeling time and improving
downstream catalog quality.
Built NLP-based pattern recognition models to generate intelligent compatibility
recommendations, increasing user engagement and conversion rates.
Designed and launched an end-to-end product recommendation workflow (data
preprocessing → embeddings → retrieval → ranking), driving measurable improvements in
CTR and session depth.
Defined LLM-in-the-loop design patterns and collaborated with Product, CX, and
Engineering teams to deliver safe, context-aware outputs in production.
Optimized prompts and model behavior through systematic fine-tuning and evaluation using
precision/recall-based benchmarks.