Best of NLP — September 2025

1. 1
   Article

   Daily Dose of Data Science | Avi Chawla | Substack·34w

   4 LLM Text Generation Strategies

   Explores four key strategies for LLM text generation: greedy selection (choosing highest probability tokens), multinomial sampling (random sampling with temperature control), beam search (maintaining multiple candidate sequences for global optimization), and contrastive search (balancing fluency with diversity by penalizing repetition). Each approach offers different trade-offs between coherence, creativity, and computational efficiency.

   16
2. 2
   Article

   habr·38w

   Building a Resume Matcher with tRPC, NLP, and Vertex AI

   A developer shares their experience building a resume matcher application using tRPC for type-safe API development, basic NLP techniques for skill extraction, and Google's Vertex AI Gemini model for intelligent matching. The project demonstrates tRPC's advantages for TypeScript-first applications, including end-to-end type safety and minimal boilerplate, while also discussing its limitations for public APIs. The implementation includes PDF text extraction, keyword extraction using natural language processing libraries, and AI-powered resume-to-job matching with structured output.

   16
3. 3
   Article

   Towards Data Science·36w

   Building Research Agents for Tech Insights

   A comprehensive guide to building specialized AI research agents that can aggregate and analyze tech content from multiple sources. The approach uses structured workflows, data caching, and prompt chaining to create personalized tech reports. Key components include preprocessing data pipelines, strategic use of small vs large language models for cost optimization, and structured JSON outputs for reliability. The system fetches trending keywords, processes facts from tech forums, and generates themed reports based on user profiles.

   12

   1