Vasanth Pugalenthi
AI Engineer and Researcher
Currently AI Engineer @ Manykind AI
CS @ Cal Poly SLO
AI engineer building and evaluating multi-agent systems. Deep learning and applied ML research background, including search and rescue work presented twice at the National Missing and Unidentified Persons Conference. Driven by building things people use, bridging research and entrepreneurship to turn novel ideas into real products.
Work
Building and evaluating multi-agent systems, with earlier work in computer vision and generative AI for healthcare.
AI Engineer
Nov 2025 - PresentShipping production agents and the evaluation harness they are scored against. Found the eval metric itself was gameable by a constant prediction, and root-caused chance-level accuracy to speaker diarization collapse rather than model quality.
Machine Learning Intern
May 2025 - Aug 2025Redesigned and fine-tuned two ResNet CNNs for real-time blurry vs sharp image classification on Zebra devices, improving accuracy from 44% to 80% on close-up images. Captured the dataset and ran the human evaluation sessions behind the ground truth baseline.
Generative AI Intern
Mar 2024 - May 2025Built conversational generative AI agents over the public ClinicalTrials.gov API that turn a natural-language question into structured queries and return grounded results, using hybrid vector and keyword retrieval.
Technical Skills
Languages
Python, TypeScript, C, SQL, Java
ML & AI
PyTorch, Scikit-learn, HuggingFace, CUDA, ResNet / CNNs, K-Nearest Neighbors, SVM, K-means
Agents & Evaluation
Multi-agent orchestration, agent SDKs, Google ADK, golden-set band scoring, human-labeled ground truth, LLM-as-judge, held-out eval sets, regression gating, automated parameter search, vitest
Speech & NLP
Speaker diarization, speech emotion recognition, acoustic feature extraction, multimodal signal fusion, hybrid retrieval (vector + keyword), RAG, conversational query generation
Infra & Web
Docker, GCP, Spring Boot, Next.js, React, Git, Linux
Coursework
Machine Learning, Deep Learning, Statistics, Linear Algebra, Calculus IV, Data Structures & Algorithms
Currently studying
Causal inference, value of information, reinforcement learning, mechanistic interpretability
Education
California Polytechnic State University
B.S., Computer Science
Merit Scholar, Cum Laude
Research
Applied machine learning research on search and rescue outcomes, plus work on GPU architectures and computational acceleration.
AI For Search and Rescue
Machine learning models achieving 80%+ accuracy in predicting missing person patterns to aid search and rescue operations.
86% recall, wander-status model
75% precision, wander-status model
2 conference presentations
15 person ML team led
Accelerating Computational Workloads: GPU Architectures, Programming Models, and Applications
Comprehensive survey on GPU-based parallel computing, exploring GPU architectures, programming models, and their applications in networking, AI, and distributed systems.
Projects
Personal and entrepreneurial projects spanning AI/ML, full-stack development, and startups.
Livin
ViewFull-stack AI-powered housing platform connecting students with sublets and roommates across Cal Poly and UC schools.
700+ users in ~4 months
Several hundred listings searchable
Cal Poly & UC schools universities
Lock In Chrome Extension
ViewProductivity Chrome extension helping users stay focused by blocking distracting websites.
Skymark Systems
ViewTech startup building innovative solutions, featured in Cal Poly CIE for turning tech vision into reality.
Writing
Thoughts on AI, research, products, and whatever else is on my mind.
Why learning backwards is better
I learn better when I start with the hard thing, see why it matters, and work backwards into the foundations.
MDPs, poker, and a $10,000 MRI
Started the day looking for a startup idea, ended up back in the foundations. Notes on decision processes, belief, and why perfect information can be worth nothing.
The thing I can't put my finger on
Started the day on routing evals, spent it on reinforcement learning instead. RL, observability, causal inference and alignment all seem to be the same question from different angles.
Why I'm starting this
A place to track what I'm reading and building while I move from the engineering side toward research.
Contact
Open to conversations about AI research, evaluation, or anything else here. Reach out.