Zara Mensah

Author

LLM product engineer at a Tema AI startup, building context-aware customer support bots with tool-use and retrieval. Reduced customer service ticket volume by 55% in the first month after deployment.

aillmapibackend

Articles by Zara Mensah(4)

Building a Semantic Changelog Diff Tool with Embeddings

Learn how we reduced support tickets by 55% using semantic changelog analysis to keep customers informed of product changes automatically.

LLMsTool-useEmbeddingsCustomer Support

3 August 2026

Structured Output from LLMs: JSON Mode vs Function Calling

Explore the benefits and applications of JSON mode and function calling for structured outputs in LLMs.

LLMStructured OutputFunction CallingJSON

31 July 2026

Building a resume parser with structured extraction

Turn unstructured resume text into clean, queryable data using LLMs with structured output. This guide covers schema design, handling inconsistent formats, confidence scoring, and building a production pipeline that processes hundreds of CVs reliably.

llmaistructured-outputbackend

19 July 2026

Fine-tuning a small LLM for domain-specific classification

Fine-tuning a small language model on domain-specific data can dramatically outperform general-purpose prompting for classification tasks while keeping inference costs low. This article covers dataset preparation, LoRA fine-tuning with the Hugging Face ecosystem, evaluation, and deployment.

fine-tuningllmclassificationpeftlora

15 July 2026