Mugil Suga Kanthimathinathan

About Mugil Suga Kanthimathinathan

I am an AI/ML Engineer and MS Computer Engineering graduate focused on building production-ready machine learning and LLM systems that move beyond prototypes and deliver measurable business impact. My work sits at the intersection of modern ML research, scalable system design, and real-world deployment.I specialize in:Designing and training transformer-based and LLM-driven systemsBuilding agentic AI applications (RAG, multi-agent orchestration, semantic search)Implementing robust MLOps pipelines to take models from experimentation to reliable productionCurrently, my work centers on agentic LLM systems and efficient AI architectures, where I combine strong theoretical foundations with hands-on engineering covering data pipelines, inference optimization, and system observability.What I have built in practice:Agentic LLM Analytics Engine converting natural-language business questions into SQL, executing queries with DuckDB, and generating visual insights end-to-end—reducing time-to-insight from minutes to ~20–30 seconds and cutting manual dashboard effort by ~70%.LLM-powered intelligent assistants using RAG, LangGraph/LangChain, and vector databases to enable semantic search over large document collections (100k+ records), with cost-aware inference strategies that reduced API spend while maintaining latency SLAs.Transformer models built from scratch, implementing multi-head attention, positional encoding, and token embeddings, then trained on curated datasets to demonstrate real-world intent recognition and response generation.End-to-end ML systems spanning data ingestion, feature pipelines, model training, experiment tracking (MLflow/DVC), containerization (Docker), and cloud deployment (AWS, CI/CD).I also bring strong experience in:Data & ML pipelines (Spark, Airflow, ETL)Production microservices (FastAPI, Flask)Model evaluation, A/B testing, and explainabilityCloud-native ML infrastructure and scalable deployment patternsCore Technical SkillsAI / ML: Transformers & LLMs, RAG, CNNs, Classical ML, Hugging Face, LangGraph, LangChainProgramming: Python (PyTorch, TensorFlow, scikit-learn), SQL, C/C++MLOps & Cloud: Docker, Kubernetes, AWS, MLflow, DVC, CI/CDSystems & Data: Distributed Systems, Microservices, PostgreSQL, Vector Databases, Apache AirflowI’m actively seeking AI/ML Engineer or Machine Learning Scientist roles. If you are building scalable, intelligent systems that blend cutting-edge algorithms with real production constraints, I would love to connect.Email: m••••••••@gmail.com

Experience

1. Ai Software Engineer

   Ephanti Inc

   Sep 2025 — Present · San Jose, CA, US

   Built an agentic LLM-powered analytics engine that converts natural-language business questions into SQL queries, executes them, and generates visual insights end-to-end with semantic search, reducing time-to-insight from minutes to 20–30 seconds.• Automated analytics workflows by orchestrating LLM-based intent extraction, query planning, execution, and visualization, eliminating manual SQL and dashboard creation and cutting dashboard development effort by 70%.• Engineered a modular Python-based analytics architecture leveraging DuckDB for query execution and PostgreSQL for metadata orchestration, achiev- ing sub-second query latency on datasets exceeding 50K records while maintaining explainable reasoning from user query to final output

Education

• BSA Crescent Institute of Science and Technology

  B.Tech Computer Science and Engineering

• The University of Texas at Dallas

  Masters in Computer Engineering