Co-founder and CEO at Daybot Solutions · University of Toronto, Computer Science 2026 · Toronto, Canada
I build products end to end: backend and distributed systems, applied ML, mathematical optimization, and the iOS apps on top of them.
Daybot Solutions runs two products with paying customers.
Daybot turns a described niche into a live, checkout ready store: supplier backed catalog, Stripe checkout, SEO pages, customer accounts, custom domains, orders and payouts.
- 124 users launched 154 stores in the first two months, with over US$7K in revenue to date, all organic with no paid acquisition.
- FastAPI on Vercel over Cloudflare D1, a Vite and React SPA, and one Expo codebase shipping both iOS and Android.
- A multi model router fans one prompt out to several LLM providers in parallel and keeps the first schema valid JSON response.
- Reviewed on YouTube by Franklin AI: I Typed One Word Into This AI and It Built the Entire Store.
RacketIQ is AI match analysis for racquet sports.
- Nothing to TestFlight in 9 days, and a squash club pilot converted 10 paid subscriptions.
- 97.9% player identification accuracy on 379 hand labelled samples; the scorer is ported to Swift so a phone referees a live match with no server.
| Project | What it is |
|---|---|
| Daybot | AI store builder, live on the iOS App Store and at daybotai.com |
| Ninja Knife Dodge | Four game offline arcade built in Godot 4, live on the App Store (repo) |
| Tarazoo | Hack the North 2025 winner, Shopify challenge. Forecasting plus SCIP purchase order optimization for merchants |
| Codexa Voice | Voice first coding agent workbench. Its phone supervisor orchestrated roughly 2,400 agent sessions in one week (demo) |
| Python to Java Translator | Fine tuned CodeT5 and PLBART, 78% compilation rate on translated output |
| ML Boilerplates | 24+ model implementations with consistent training and evaluation loops |
| Matroid Research | Polytope volume estimation under Prof. Ahmed Ashraf |
- IPPINKA, Machine Learning Engineer and E-Commerce Architect (2024 to 2025). Event driven inventory and order sync on GCP across 5,000+ SKUs, demand forecasting that lifted inventory turnover 23%, and a Pyomo purchase order optimizer over SCIP and CPLEX.
- KPMG, AI/ML Engineer, Data Governance (2023). LLM assistants that cut manual compliance review 60%, PII detection architectures on GCP and Azure, and the firm's Data Governance Playbook.
Languages Python, TypeScript, Swift, SQL, C/C++, Java, Go, GDScript, Bash ML and AI PyTorch, Hugging Face, scikit-learn, XGBoost, Prophet, OpenCV, RAG and embeddings, LLM APIs Backend FastAPI, Flask, Node, REST and WebSockets, PostgreSQL, Cloudflare D1, Redis, Cassandra, BigQuery Frontend and mobile React, Vite, Tailwind, Expo and React Native, SwiftUI, Godot 4 Cloud GCP, AWS, Azure, Vercel, Cloudflare, Docker, Kubernetes, Terraform, GitHub Actions Optimization SCIP, CPLEX, Gurobi, Pyomo, MILP and MINLP Agentic development Claude Code, Codex CLI, MCP servers, multi agent orchestration
Hack the North 2025 winner (Shopify challenge) · Entrepreneur First Selection Day, San Francisco 2026 · Y Combinator founder dinner invite 2025 · Halton Game Jam 2025 winner · UofT Hacks 2023 runner up · DeerHacks 2022 winner · Dean's List 2021 to 2026 · DAMA CDMP Associate
kkaranvir.me is my portfolio, built as an interactive desktop.