VADLAKONDA CHARAN KUMAR

Performed prompt engineering and structured QA reviews on large language model outputs by evaluating accuracy, coherence, safety, and instruction-following. Designed and refined prompts against defined evaluation guidelines, and applied hallucination detection, factual error identification, and policy violation checks. Collaborated with cross-functional AI training teams to document edge cases, calibrate evaluation rubrics, and maintain data quality for downstream fine-tuning cycles.

Designed and deployed end-to-end machine learning systems and LLM-powered applications using Python, FastAPI, and Docker to achieve sub-200ms API latency in containerized MLOps environments. Engineered RAG pipelines integrating FAISS and ChromaDB across structured/unstructured data to improve query accuracy by 30%. Built deterministic LLM evaluation frameworks for hallucination rate, content safety, bias detection, and conversational quality, enabling automated benchmarking. Automated agentic workflows using LangChain and LangGraph, reducing manual data operations by 35%, and containerized ML workloads with Docker. Built SQL ETL pipelines with data validation and integrity checks for downstream training workflows and produced bilingual Power BI dashboards for KPI tracking and faster decision-making.

Solved 200+ complex problems in machine learning, deep learning, NLP, algorithms, data structures, and Python programming while maintaining high accuracy. Reinforced conceptual clarity in AI/ML topics including supervised/unsupervised learning, model evaluation, NLP, and statistical analysis for a global learner base.