AI Engineer · Tempe, Arizona
Building production-grade Generative AI: RAG pipelines, multi-agent architectures, and real-time ML systems that scale.
// about me
I'm an AI Engineer pursuing my Master's in Computer Science at Arizona State University (GPA 3.89), specializing in the full stack of modern Generative AI, from vector retrieval and agent reasoning to cloud deployment and observability.
My work spans multi-modal emotion recognition, agentic voice pipelines, MLOps platforms, and RAG hallucination detection. I care deeply about systems that are reliable, measurable, and actually ship.
Currently building at ASU's RVCoLab MIX Center, where I design GenAI pipelines that power AI-driven virtual avatars in real time.
// technical skills
// experience
// projects
Full-stack MLOps platform for personalized habit optimization. Contextual multi-armed bandit system (Thompson Sampling, UCB, Epsilon-Greedy, LinUCB) using 15 behavioral signals over 30-day windows to dynamically select from 10 nudge types.
RAG firewall estimating answer reliability before responses reach users — combining semantic entropy, Jensen-Shannon divergence, and DeBERTa NLI contradiction scoring into a per-query hallucination risk score. Benchmarked on 198 research papers (1,225 chunks, Pinecone with MMR retrieval) at sub-200ms latency.
Multi-agent LLM system with ReAct-based reasoning loops, tool integration (search, code execution, file I/O), and persistent vector memory for multi-session context. Reduced manual review effort by 30–40%.
Speech command recognition for 35 voice commands trained on 105K audio samples — achieved 96.27% accuracy and 11.4ms inference latency by selecting a CNN-BiLSTM-Attention architecture over 4 alternatives. Packaged into a production inference service with 40-dim MFCC preprocessing, SpecAugment, and Gaussian noise augmentation.
MobileNetV2 trained on 12K dermoscopy images across 8 disease classes with 95% diagnostic accuracy. Co-authored a granted patent (IND 102069, Sept 2022). MC Dropout for uncertainty estimation improved generalizability by 18%.
// education
// contact
Open to AI/ML engineering roles, research collaborations, and interesting conversations about agentic systems.
sgudiva3@asu.edu