Agent Nomad
Built a belief-aware AI agent marketplace where agents bid on tasks based on confidence, with live analytics through AutoDB + Postgres and research workflows powered by Claude and Tavily.
A collection of my work spanning web development, full-stack applications, ML systems, and hackathon experiments.
ASU AIR Spark • First Place
Reads every picture in a PowerPoint lecture deck and writes a WCAG description back into the file, so a blind student's screen reader can explain diagrams instead of saying "image." A confidence gate sends uncertain images to human review instead of guessing. It described 380 images across 505 slides in under 10 minutes. Won First Place and Crowd Favorite at the ASU AIR Spark Challenge 2026.
Kiro Spark • Best Collab Winner
Built a voice-first AI conference companion that captures recruiter conversations, scores connection value transparently, and drafts context-aware follow-ups. Won Best Collab at Kiro Spark Challenge 2026.
Created a voice-enabled AI onboarding tutor for real estate agents, using live DOM awareness, grounded Gemini responses, ElevenLabs speech, and Insforge edge functions for interactive CRM walkthroughs.
Built a belief-aware AI agent marketplace where agents bid on tasks based on confidence, with live analytics through AutoDB + Postgres and research workflows powered by Claude and Tavily.
Hackathon • Developer
ASU WiCS Opportunity Hack project - A collaborative platform for solving problems together in real-time with team synchronization features.
Hackathon • Developer
Financial clarity tool helping users understand and manage their finances with clear visualizations and insights.
AI SaaS • Developer
Built an AI research workspace that turns uploaded papers into summaries, insights, implementation guidance, grounded Q&A, and a runnable Python sandbox.
AI + Data • Developer
Created a real-time emergency response command system that ingests live incident feeds, scores operational risk, gates AI triggers, and serves dispatch-ready insights through an interactive map.
Backend • Lead Developer
Designed a real-time graph analytics backend where events are ingested via Kafka, intermediate state is stored in RocksDB, and graph algorithms run in Neo4j on Minikube.
ML • Developer
Built an ML clinical decision system to predict hospital readmissions in diabetic patients using EHR data. Deployed an interactive Streamlit dashboard using SHAP for local explanations.
ML • Developer
Designed an end-to-end ML pipeline to detect plant leaf diseases using segmentation, color histograms, and Haralick textures. Achieved 97.5% accuracy using Random Forest.
NLP • Researcher
Performed sentiment classification on Hinglish code-mixed datasets using MLP and CNN models. Applied TF-IDF and Word2Vec embeddings alongside emoji normalization.
Full-Stack • Lead Developer
Built and deployed a full-stack account trading platform, boosting engagement by 20%. Integrated secure authentication and role-based access using JWT and bcrypt.
Full-Stack • Developer
A real-time platform relying on anonymous device hashes to passively track and compute crowd densities. Engineered a decay-based aggregation algorithm.