Animesh Kumar 7459958032 | animesh.kr9044@gmail.com | www.linkedin.com/in/iamanimesh | github.com/Animeshkr9044 Education Harcourt Butler Technical University B.Tech in Electrical Engineering Kanpur, Uttar Pradesh July 2020 – July 2024 Experience Machine Learning Engineer July 2024 – Present Deccan AI Hyderabad • AI Interviewer: Developed an end-to-end AI interviewer using Retell.ai, creating a fully customized system to conduct resume-based interviews. Built the entire infrastructure, including backend (Node.js), frontend (React.js), and deployed it using AWS EC2 instances. • Text2SQL Benchmark: Conducted a comprehensive evaluation of top-performing models and frameworks for Text2SQL. Compared open-source datasets such as Spider and Bird Dataset with in-house datasets to assess performance. Developed two dedicated pipelines: one for generating Text2SQL queries and another for obtaining performance metrics. • Conducted a thorough literature survey and selected the top-performing Text2SQL framework (ChaseSQL), and successfully implemented it for query generation and evaluation. • LLM-as-Judge: Implemented LLM-based automated quality checks across multiple projects to enhance accuracy and efficiency. Developed an automated quality control framework for Text2SQL, significantly reducing workload by efficiently detecting errors. Additionally, built an automated QC process for Multi-Modal Data Interpretation, improving data validation accuracy and ensuring consistency across diverse data sources. Software Engineer Intern Dec 2023 – July 2024 Callmatic AI Remote • Developed a RAG (Retrieval-Augmented Generation) system for voice agents by integrating MongoDB vector database and Pinecone for efficient data storage and retrieval. • Enhanced the knowledge base functionality in Vocode, contributing production-ready code to optimize voice-based AI interactions. • Identified and resolved latency issues within the open-source Vocode framework, successfully reducing response time from 2 seconds to 50 milliseconds by leveraging smaller models. • Deployed an automated script on Google Cloud Platform (GCP) to process PDF documents via an event-driven mechanism, enabling seamless knowledge base updates. Projects Finance RAG Quality Check Framework | Python, PyPDF, GPT-4, NLP 2024 – Present • Developed a framework to ensure the accuracy and coherence of LLM-generated responses in finance by implementing similarity scoring and citation validation. • Built automated citation verification and linguistic quality checks to validate page numbers, headers, and grammar with a text-based search approach. • Designed a system to analyze sub-questions, count citations, and automate workflow using Python. AI-Powered Search Response | Google Custom Search API,Anthropic Claude 2024 – Present • Developed an automated framework integrating Google Custom Search API with OpenAI GPT-3.5 and Anthropic Claude to generate AI-driven responses based on real-time web search results. • Implemented a robust pipeline to process search queries efficiently and store responses for further analysis. • Enhanced response quality by incorporating citation validation and hyperlinking in AI-generated responses to ensure accuracy and reliability. Technical Skills Programming Languages | Python, Bash, JavaScript Frameworks and Libraries | FastAPI, Transformers, Streamlit, Gradio, TensorFlow, LangChain, Crew.ai Cloud Platforms | AWS (EC2), Replit Machine Learning and AI | LLMs, Retrieval-Augmented Generation (RAG), Prompt Engineering,Fine-Tuning