Headline

AI/ML Engineer | Production RAG & Real-Time ML Systems | LangChain · FastAPI · PyTorch · Python | Open to work

About

I build production-grade GenAI and ML systems focused on retrieval, reasoning, and real-time inference. At ASU's RVCoLab, I work on multimodal AI systems combining vision, biosignals, and LLMs to drive real-time behavior in interactive agents. Key projects: Built a RAG hallucination detection system across 198 arXiv papers with sub-200ms latency using semantic entropy, divergence metrics, and NLI-based scoring. Developed production ML pipelines for 3D-printed metal property prediction (92% accuracy), deployed via FastAPI. Trained and benchmarked deep speech recognition models across 105K samples -- best model achieved 96.27% validation accuracy at 11.4ms inference latency. Delivered RAG systems used by 170+ users across 1,800+ enterprise documents at ~3s response time. Tech: LangChain · OpenAI · PyTorch/TensorFlow · FastAPI · Pinecone · Chroma · AWS · Azure · GCP · MLflow MS in Computer Science, Arizona State University (GPA: 3.87, Dec 2025). Actively seeking GenAI and ML Engineer roles building scalable AI systems. If you're working on something ambitious with LLMs or real-time AI, let's connect.

Location

Kolkata

AI stack

deepseek

hugging face

langchain

langgraph

langsmith

mcp

Work experience

AI & Data Engineer at ASU Media and Immersive eXperience (MIX) Center, Feb 2026 - Present · 10 mos, Architecting a real-time multimodal data pipeline synchronizing 4 concurrent streams (facial capture, biometric signals, Discord logs, gameplay footage) with frame-accurate timestamp alignment. Building Python middleware fusing live biosignal and facial expression data to drive adaptive behavior in an Unreal Engine 5 avatar, targeting sub-50ms response latency. Part of ASU's Resilient Visions Collaboratory Lab (RVCoLab) research on AI-mediated emotional engagement.

AI / Software Engineer Intern (Applied AI) at TechieGenz Softwares, Nov 2025 - May 2026 · 7 mos, Built and shipped 2 production RAG deployments processing 1,800+ enterprise documents, serving 170+ users at ~3s end-to-end response latency. Automated 2 manual document-processing workflows using FastAPI-based AI services, reducing review time by 65%. Built 4 RESTful APIs for document querying, summarization, and chatbot functionality. Improved retrieval precision for multi-document semantic search by replacing keyword matching with vector embeddings and OpenAI APIs, validated through A/B testing.

Grader at Ira A. Fulton Schools of Engineering at Arizona State University, Aug 2025 - Dec 2025 · 5 mos

Graduate Service Assistant at Ira A. Fulton Schools of Engineering at Arizona State University, Sep 2024 - Aug 2025 · 1 yr, Tempe, AZ, – Preprocessed 5,000+ microscopy images using OpenCV and NumPy (denoising, normalization, patch extraction), achieving 25% noise reduction; structured cleaned dataset into a versioned data pipeline with MLflow experiment tracking. – Applied CNNs and statistical feature extraction to microstructure classification models, boosting prediction accuracy by 16%; used GPT-4 with RAG over research PDFs to surface domain-specific augmentation strategies that improved model generalizability. – Delivered a production-ready Python ML pipeline (TensorFlow, Scikit-learn, Pandas) achieving 92% accuracy for predicting mechanical properties of 3D-printed metals, with a FastAPI inference endpoint for researcher access.

Artificial Intelligence Intern at Planet E-Com Solutions (PECS), May 2022 - Jun 2022 · 2 mos, • Created an AI-driven program aimed at facilitating student overseas travel by reducing consultation costs and streamlining university application processes, resulting in an 13% increase in efficiency. • Applied machine learning to analyze SAT and TOEFL scores, delivering personalized English and math course recommendations from top universities, boosting student satisfaction and success. • Developed an AI-powered interface to guide students through international education, aiding in form completion, policy understanding, and university selection for a streamlined decision-making process.

Machine Learning Intern at SmartKnower, Aug 2020 - Sep 2020 · 2 mos, • Developed machine learning algorithms for data analysis, including supervised and unsupervised learning techniques, resulting in a 10% improvement in predictive accuracy compared to previous methods. • Collaborated with cross-functional teams to gather requirements, conduct data preprocessing, and refine model performance, demonstrating strong communication and teamwork skills in a fast-paced environment. • Established a scalable data pipeline using tools such as TensorFlow and PyTorch, optimizing model training time by implementing parallel processing techniques and leveraging cloud computing resources.

Education

Master's degree, Computer Science, Arizona State University

Bachelor of Technology - BTech, Artificial Intelligence, amity University

Public links

github: https://github.com/swathi-2406