Backend developer

Koshal Goyal  • Peerlist

A small question-answering service over a tiny, real document set. First ship a basic similarity search. Then make it smarter with a reranker so better evidence rises to the top. Finally, show a simple before/after comparison to prove the improvement

Intelligent Safety Docs Search with RAG + Cus

• Designed and implemented a production-ready, real-time document processing pipeline using Go, Kafka, and Couchbase, enabling scalable ingestion, enrichment, and full-text indexing of business documents, emails, and logs. • Developed microservices for content extraction, NLP enrichment (TF-IDF keywording, classification, summarization, sentiment, entity recognition), and search, with robust error handling, health checks, and observability for reliable operations. • Built custom NLP modules without LLMs to avoid trivial API-based solutions, focusing on classical techniques for language detection, keyword extraction, and document classification to ensure transparency, efficiency, and production control.

real-time-doc-indexing

• Engineered an AI-powered web crawler to scrape structured content from dynamic websites using user-defined prompts; designed for seamless integration with RAG pipelines and LLM-based systems. • Built end-to-end backend (FastAPI, Playwright, LangChain) including dynamic content handling, intelligent chunking, and prompt-based routing with Gemini API. • Achieved efficient multi-domain data extraction with automated error handling, enabling applications in chatbots, auto-documentation, and LLM fine-tuning.