Nikhil Singh

About Nikhil Singh

Hi, I’m Nikhil — an AI and Data Science professional with 5+ years of experience working across Generative AI, machine learning, and statistics to turn complex data into actionable business insights.My recent work focuses on GenAI applications, language assessment and grading, advanced multi-agent systems, and large language models, particularly in the education and evaluation space. I enjoy building solutions that are not only technically strong but also practical and impactful.I’ve been fortunate to receive recognition for my work, including: Winner of the UK Data IQ Award (Transformation in AI)🥈 National-level finalist at the ULIP Hackathon, organised by the Ministry of Commerce and IndustryI’m driven by solving meaningful problems using Data Science and AI, whether that’s through predictive modeling, intelligent systems, or applied GenAI solutions.I’m always happy to connect—especially with students and early-career professionals seeking guidance or mentorship. If you’re interested in AI, data, or building real-world solutions, feel free to reach out.

Experience

1. Senior Data Scientist

   British Council

   Oct 2023 — Present

   Developed a RAG model for summarizing large reports, combining vector search with Naive and Knowledge Graph RAG. Implemented a Self-Agentic RAG architecture with intelligent query refinement, metadata-aware retrieval, and automated fact verification. Built a Query Refinement and Fact-Checking Agent to enhance accuracy by iteratively improving queries and validating retrieved content against metadata. Evaluated model performance using REHL, RAGAs, and DEEPeva metrics. Led the creation of an AI-driven scoring system for the APTIS for Teens writing task using ML and LLMs. Utilized Langchain for feature generation and custom word lists, achieving a 25% improvement in scoring precision. Developed and implemented an Aspect-Based Sentiment Analysis (ABSA) project to analyze social media opinions about the British Council and UK universities using models like VADER, BERT, RoBERTa, and SLMs. Optimized token limits through key phrase extraction and applied topic modeling to identify drivers of sentiment, validated by human scoring.

Education

• Aligarh Muslim University

  Master's degree