TRIBUNAL: A Machine Court for Production Incidents
GitHub: https://github.com/UjwalKandi/tribunal
- Built an adversarial AI court for production incidents, where Prosecution, Defense, and Judge LLMs argue each fix before a ruling.
- Rulings pass a Flink-enforced human veto window, then become Kafka precedents that future hearings must argue against.
- Developed at the Cursor Austin × AITX Hackathon as a successor to A.I.D.E., using Next.js, TypeScript, Claude structured outputs, Zod, and Supabase pgvector, with cases drawn from real Airflow and dbt failures.
FlipFinder Pro: Real-Time AI for Car-Flip Profit Detection
GitHub: https://github.com/UjwalKandi/FlipFinder-Pro
- Built for the Databricks Free Edition Hackathon — an ML and anomaly-detection model that analyzes used-car data, predicts market values with 89%+ accuracy, and surfaces high-profit flip opportunities.
- Developed an interactive analytics dashboard that reveals flip scores, risk levels, profit potential, and $20M+ worth of opportunities.
A.I.D.E. - AI Diagnostic Engine
GitHub: https://github.com/UjwalKandi/A.I.D.E
- Built an AI tool to analyze logs locally, identify issues, and provide diagnostic reports.
- Utilized Llama models on Groq, implemented RAG with ChromaDB, and created visualizations with SciPy for enhanced analysis.
- Developed the initial prototype at the ATX Llama Hackathon hosted by Meta, where I used the Llama Stack for on-device intelligence.
Customer Characterization & Profiling using Agglomerative Clustering
- Conducted comprehensive analysis of client habits, demographics, and marketing responses using different clustering techniques.
- Opted for Agglomerative Hierarchical Clustering, gaining deeper insights into distinctive customer segments.
Smart Air Quality Monitoring: IoT Sensors on M5StickC Plus
- Developed a Smart Air Quality Monitoring System using M5StickC Plus, AT6558 GPS, and SGP30 Gas Sensor, seamlessly integrated with Qubitro.
- Conducted live demonstrations with threshold alerts, showcasing the system’s ability to detect harmful gases.
Archive - Photography Portfolio
https://ujwalkandi.in/archive/
A curated collection of moments captured through my lens. Dive into a world of vibrant landscapes and sleek rides, each shot infused with the energy and chaos of my imagination.
ujwalkandi.in
(the website you are currently on)
- Deployed website to IPFS using Fleek: https://black-frog-5099.on.fleek.co/
- Uses JAM Stack - Javascript APIs Markup.
- Uses Hugo (theme PaperMod) to render Markdown pages.
- Vercel’s Serverless for APIs and loading badges with image-proxy.
- For analytics I use Panelbear.com.
EnDe-K
GitHub: https://github.com/UjwalKandi/EnDek
- Python code for implementing Encrypting File System (EFS) using Modular Key framework.
Dashboard on Drug Cartels' Activity across Mexico
Tableau: https://public.tableau.com/app/profile/ujwal.kandi/viz/DrugCartels-Mexico/DrugCartels-Mehico
- Extracted data of 10 drug cartel’s drug traffic activity spanning over
20years. - Analyzed shifts in their territory and impact over each state from 1989 to 2010.
Financial Impact of Presidential Tweets on Stock Markets
Tableau: https://public.tableau.com/app/profile/ujwal.kandi
Correlation Viz
- Scrapped vital tweets of world leaders from a Tweet Archive and analyzed their financial impact on global stock markets.
- Applied several Data Visualization techniques to indicate the correlation between the two factors.
Archaeological Sites with Mayan Inscriptions
Tableau: https://public.tableau.com/app/profile/ujwal.kandi/viz/MayanSites/MayanSites
- Visualized a dataset of
334newly discovered sites across Mesoamerica that expands the known territory of the Mayan civilization. - Mapped lat-long coordinates to sites that lacked GPS values through geocoding.
Real-Time Face Recognition with Raspberry Pi 3
GitHub: https://github.com/UjwalKandi/Object-Recognition-with-Raspberry-Pi
- Developed model using OpenCV, Numpy, Dlib and implemented with Raspberry Pi 3.
- Provides real-time confidence score of the faces recognized.
Deaths during the LA Riots of 1992
Tableau: https://public.tableau.com/app/profile/ujwal.kandi/viz/LAriots1992/DeathsduringLAriotsin1992
- Acquired statistical data on homicides & other riot-related deaths from the official investigation records of LAPD, LASD, and LA Times.
- Also gathered data from the building records of LADBS on properties that were destroyed by arson during the riots.
the Note Making Project 📝
GitHub: https://github.com/UjwalKandi/the-Note-Making-Project
- Implemented Optical Character Recognition using MATLAB as part of the Note Making Project.