I'm Likhith Reddy Bandi, AI Engineer at Meta. I specialize in building algorithms for high-precision optical metrology, and I love coding models from scratch to understand them under the hood.”
AI/ML Engineer driven by the challenge of taking LLMs, AI concepts out of the lab and into production. Specialize in designing scalable AI architecture, optimizing model inference, RAG architecture, multi-agent orchestration, model fine-tuning, building end-to-end production pipelines. Experienced in translating applied AI research into reliable, high-throughput production systems, and boosting inference efficiency at scale.
Strong skills in building scalable data pipelines, ETL/ELT pipelines, data analysis, statistical analysis, data visualization, SQL, Python, and EDA to solve complex business problems. Proficient in managing large-scale datasets, automating reporting workflows, ensuring data quality, identifying trends and delivering actionable insights through data-driven decision-making.
Jan 2025 - Present
Meta
Designed Meta AI RAG platform using FAISS, Llama, hybrid retrieval, REFRAG-inspired compression, and LLM evaluation, improving latency, retrieval quality, and scalability.
May 2022 – December 2023
IQVIA
Developed healthcare GenAI applications using RAG, LLMs, Pinecone, AWS, and Vision-Language Models, automating claims review, knowledge retrieval workflows and reduce processing time.
MEESHO
Developed recommendation and search systems using XGBoost, NLP embeddings, click, browsing, purchases, and product metadata enhancing improving personalization, product discovery and search relevance.
2024 -2025
The University of Texas at Dallas
2019-2023
National Institute of Technology
Retrieval-Augmented Generation (RAG) application that combines document management with AI-powered question answering.
Deployed Brand Guardian, a multimodal agentic AI platform that reduced manual video compliance review time by 70% via a multimodal agentic pipeline that cross-referenced video content against brand/regulatory guidelines in real time, with full observability stack for production reliability.
A powerful web application that allows you to upload Excel files and ask natural language questions to generate SQL queries and get data insights. Built with Streamlit and OpenAI's GPT models
This project focuses on applying natural language processing (NLP) techniques for emotion classification using advanced machine learning models. A variety of models, including LSTM, Transformer-based models (RoBERTa, DistilBERT, ALBERT), QLoRA models (SFR, GEMMA, LLaMA), and fine-tuned transformer models, were implemented and evaluated to identify the most effective approach.
This is a category-aware Conversational AI chatbot developed for the Jindal School of Management (JSOM) at the University of Texas at Dallas. It allows users to ask natural language questions and receive accurate, relevant answers drawn from structured website content.