Sanjay Karinje

Sanjay Karinje

Full-Stack AI Engineer • Data Scientist

Built ML systems that drove $300M+ impact at Amazon & Instacart. Stanford AI, Gauntlet Fellow, UF MBA. Currently building Metriq.

Selected Projects

A curated list of my recent work.

Weftly

WhatsApp-style iOS messenger with AI agent for busy parents

iOS messaging app built for busy parents drowning in group chats. A unified AI agent automatically extracts calendar events, deadlines, RSVPs, and priority messages from conversations. Features conflict detection against device calendar and semantic search via RAG.

SwiftSwiftUISwiftDataFirebaseLangGraphPinecone

CanvasIQ

Figma-meets-ChatGPT collaborative whiteboard with voice/text commands

Real-time collaborative whiteboard with AI-powered canvas manipulation. Users can create and edit shapes via natural language commands through a LangGraph-based agent with 19 specialized tools. Supports 600+ objects at 60 FPS with sub-100ms sync.

ReactTypeScriptKonva.jsFirebaseLangGraphGPT-4o

Game Eval

Autonomous QA agent that plays and scores browser games using LLMs

Autonomous AI-powered game testing platform using browser automation and LLMs to evaluate browser-based games through a multi-phase pipeline: load validation, control discovery, gameplay exploration, and quality scoring.

Node.jsTypeScriptBrowserbaseOpenAI API

Step Tutor

AI math tutor that teaches through questions, not answers

AI math tutor that guides students through problems using the Socratic method, asking questions instead of giving answers. Integrates with GeoGebra canvas for visualizing geometry and functions through a 3-node agent architecture with 20 drawing tools.

Next.jsTypeScriptOpenAI APIGeoGebra

Clip Forge

Desktop video editor with screen recording and PiP composition

Desktop video editor with screen recording, webcam capture, and Picture-in-Picture composition. Features multi-track timeline editing, professional keyboard shortcuts, and FFmpeg-powered export to MP4/WebM/MOV at up to 1080p.

ElectronReactTypeScriptZustandFFmpeg

SpendSense

AI financial coach that learns your spending habits and educates

Consent-aware financial education platform that syncs transaction data from banks, computes behavioral features, assigns user personas, and generates personalized education content. Multi-layer guardrails ensure recommendations are safe and appropriate.

PythonFastAPIReactTypeScriptFirebase

Metriq

Cursor for data scientists and analysts

Private

AI-native IDE built on JupyterLab with MCP tools and LangGraph agents. Integrates databases, dashboards, and external systems, eliminating fragmented context across tools and turning analyses into repeatable recipes with memory.

PythonJupyterLabLangGraphMCPReact

Experience

  1. Founder / Software Engineer

    Metriq San Francisco, CA

    • Metriq: Cursor for Data. AI-native IDE (JupyterLab fork + MCP tools + LangGraph agents) that integrates databases, dashboards and external systems, turning analyses into repeatable recipes with memory.
  2. Sr Data Scientist II, Membership Growth

    Instacart San Francisco, CA

    • Led Data Science for Instacart+ growth & retention: Partnered with PM/Eng to launch member benefits that drove >$300M incremental revenue.
    • Built incrementality framework for the membership program (long-term holdout + propensity score matching + regression), informing ~30% higher investment.
    • Diagnosed churn and recommended product fixes that improved member retention ~8%.
  3. Data Science Lead, Alexa

    Amazon Seattle, WA

    • Designed experiments and developed ML models to improve quality of human generated labels that train Alexa's neural networks.
    • Developed tree ensemble/deep learning models to predict errors in human labeled transcriptions, achieving 20% quality improvement by reprocessing only 5% of utterances.
  4. Data Analytics Manager

    Amazon Seattle, WA

    • Hired and managed a team of 15 analytics professionals focused on optimization of Amazon's outbound shipping cost, delivering >$70M/year cost savings.
    • Established metrics for 'Amazon Air' program during launch/incubation. Developed ML model that identified system misalignment, resulting in $25M/year savings.
  5. Financial/Senior Financial Analyst, Retail

    Amazon Seattle, WA

    • Primary finance owner of Automotive product line with full P&L ownership (>$XB+ in Product Revenue).
    • Proposed inventory system changes that reduced inventory by ~4% across all product lines in the Hardlines vertical.
  6. Design Engineer, Core Cell Libraries

    Texas Instruments Bangalore, India

    • Designed and optimized standard cell libraries for advanced CMOS process nodes.

Education

Gauntlet

Oct 2025 - Dec 2025

Gauntlet AI Fellow

Highly selective program focused on production AI engineering: RAG pipelines, AI agents, multi-agent systems, fine-tuning, and RL.

Stanford University

2023 - 2024

AI Graduate Certificate

CS231n, CS224n, CS229, CS224w • GRE 340/340

University of Florida

2010 - 2011

Master of Business Administration

GPA 3.8/4.0 (top 10%)

PES Institute of Technology

2002 - 2006

Bachelor of Engineering

Ranked 270/100k on entrance examination