Open to opportunities

ISHITA SINGH

@ishitasingh

Message

I build and evaluate grounded RAG and multimodal LLM applications.

India

Message

What I'm looking for

I'm looking to build grounded GenAI applications and RAG systems, with opportunities to improve LLM evaluation, hallucination detection, monitoring, and human-in-the-loop workflows.

I've evaluated multimodal LLM outputs at Ethara AI, building hallucination-detection frameworks across text, audio, and video models. My work flagged 200+ hallucination instances for retraining and human-in-the-loop review pipelines.

I established monitoring and error categorization across 500+ prompt-response pairs, identifying failure patterns in complex multimodal data streams. The resulting reports reduced recurring system errors by 30–40% within the first month.

I also built DocuBot, an end-to-end document Q&A RAG pipeline using Python, LangChain, HuggingFace, FAISS, Streamlit, and Groq, with sub-three-second query responses and source-grounded prompting.

Alongside GenAI work, I developed ConstructPro for the Government of Bihar and published research on a hybrid quantum machine learning model for protein-folding prediction.

Experience

Work history, roles, and key accomplishments

LLM Post-Training Evaluator

Ethara AI

Feb 2026 - May 2026 (3 months)

Built and executed an evaluation framework for hallucination detection across 3+ LLM models spanning text, audio, and video modalities, flagging 200+ hallucination instances integrated into retraining and human-in-the-loop review pipelines. Established a structured monitoring and error-categorization process across 500+ prompt-response pairs, analyzing failure patterns across complex, multi-modal

LLM Evaluation Hallucination Detection Monitoring Human In The Loop Retraining Pipelines

Frontend Developer Intern

Government of Bihar

May 2024 - Jun 2024 (1 month)

Built ConstructPro, a construction-management web application with interactive data visualizations for budget, labor, and material tracking, serving 20+ government officials. Implemented session-based authentication with role-level access control.

Frontend Development Web Applications Data Visualization Authentication Role Based Access Control