Currently Open to Work • New York, NY

Hello! I'm
Atharva,
Full-Stack
AI Product Builder

I ship AI-powered web apps from whiteboard to production — fast. 3+ years building GenAI systems, automations, and React/TypeScript tools using Claude and Cursor daily. Not notebooks. Real apps. Real users.

View Projects → Download Resume ↓
Claude API • MCP Next.js • TypeScript RAG • Production
Atharva Deshmukh — Full-Stack AI Product Engineer based in New York, NY. Builds and ships production web applications using React, Next.js, TypeScript, FastAPI, Supabase, and the Claude API.
Atharva D.
AI Product Engineer
CVS Health • CrowdDoing • VR Digital Solutions • Next.js 14 • Claude API • Supabase • Vercel • Tailwind CSS • AWS SageMaker • Hugging Face • LangChain • Microsoft Azure • Railway • Oracle Cloud • CVS Health • CrowdDoing • VR Digital Solutions • Next.js 14 • Claude API • Supabase • Vercel • Tailwind CSS • AWS SageMaker • Hugging Face • LangChain • Microsoft Azure • Railway • Oracle Cloud •
3+
Years Experience
Building production ML & GenAI systems
10+
Projects Delivered
From MVPs to enterprise platforms
2
Cloud Certifications
Azure & Oracle Cloud Data Science

About Me

Building AI That
Actually Scales

I'm a full-stack AI product engineer who builds and ships real web applications — not just models. Give me a vague idea and I'll turn it into working software. I use Claude and Cursor every day to move faster than traditionally possible.

3+ years shipping production GenAI systems, React/TypeScript apps, automations, and internal tools across healthcare and enterprise. My stack: Next.js, TypeScript, FastAPI, Supabase, Claude API, and whatever else the job needs. Currently completing my Master's in IT & Analytics at RIT.

MS
IT & Analytics
RIT, 2025
BE
Computer Engineering
Pune University
Data Science
Statistical analysis, predictive modeling, and data visualization expertise
Machine Learning
Deep learning, neural networks, and ensemble methods for complex problems
Generative AI
LLMs, RAG systems, MCP, and multi-agent AI architectures
Product Engineering
Full-stack web apps, Claude API integrations, and automations — shipped to production fast

Professional Journey

Where I've Worked

Forward Deployed AI Engineer (Contract)

ContextQA Feb 2026 – Mar 2026
  • ►Built React + TypeScript CRM to capture and manage 115+ AI-sourced leads, reducing sales rep processing time by 60% through automated pipeline management.
  • ►Delivered 18 qualified prospects (16% conversion rate) within 2 weeks, validating an agentic outreach pipeline from POC to production.
React TypeScript Agentic AI CRM

Data Scientist (AI/ML)

CrowdDoing Sep 2025 – Present
  • ►Built wildfire satellite image-classification model using PyTorch (Landsat-8 + MODIS), achieving 87% precision and 82% recall, reducing false alerts by 30% across monitored regions.
  • ►Developed real-time inference pipeline on AWS SageMaker, combining environmental sensor signals with ensemble ML models to prevent early wildfire spread.
  • ►Integrated LLM-based natural language search (LangChain + OpenAI) into dashboards for conversational access to risk insights, improving coordination efficiency by 25%.
  • ►Built MCP-powered AI agent for automated QA with council-of-agents validation, dynamically generating Playwright tests, reducing manual testing time by 70% while improving coverage to 95%.
PyTorch AWS SageMaker LangChain MCP Playwright

GenAI Engineer

CVS Health Jan 2025 – Feb 2026
  • ►Built clinical RAG pipeline using pgvector and Azure AI Foundry, enabling real-time retrieval from 2M+ patient records with RAGAS-evaluated answer accuracy.
  • ►Fine-tuned clinical embeddings with QLoRA (Hugging Face) to improve medical case-classification accuracy by 19%, enabling more reliable treatment prioritization.
  • ►Deployed FastAPI inference microservices on GCP Vertex AI serving 10K+ daily clinical predictions for pharmacy and case-management teams.
  • ►Implemented MLflow experiment tracking with CI/CD integration, reducing release cycle time by 30%.
RAG pgvector Azure AI Foundry QLoRA RAGAS FastAPI

ML Engineer

VR Digital Solutions Mar 2021 – Jul 2023
  • ►Built XGBoost-based customer segmentation and churn models on 2M+ records, improving marketing ROI by 17% across multiple client accounts.
  • ►Deployed YOLO-based defect detection pipeline for product image QA, saving $200K+ in manual inspection costs and reducing defect escape rate by 35%.
  • ►Engineered ETL workflows using Python, SQL, Pandas, and Airflow to process 3M–8M daily records, improving pipeline throughput by 2.1x.
  • ►Standardized MLflow experiment tracking and model registry across team, enabling reproducible deployments and cutting release cycle time by 25%.
XGBoost YOLO PyTorch MLflow Airflow

Technical Skills

My AI & Data Toolbox

A comprehensive suite of tools and frameworks I use to design, build, and deploy AI systems at scale.

Programming
Python SQL R Java TypeScript
ML / DL Frameworks
TensorFlow PyTorch Scikit-learn XGBoost YOLO
Cloud Platforms
AWS Azure AI Foundry GCP Vertex AI OCI
Gen AI & LLMs
Claude API LangChain LangGraph CrewAI n8n RAG QLoRA RAGAS MCP
MLOps & DevOps
Docker Kubernetes Terraform MLflow Airflow CI/CD Observability
Product Stack
Next.js 14 React Tailwind CSS Supabase Vercel Railway FastAPI
Data & Databases
PostgreSQL pgvector Redis FAISS Pinecone Pandas Tableau

Featured Work

Projects That Ship

Published Plugin ↗
Sloppy-Stoppie Live
Claude Code plugin published to the Claude marketplace — enforces design discipline by detecting and flagging AI-generated UI anti-patterns. 2★ on GitHub.
JavaScript Claude API Playwright
Agentic AI ↗
Vizard AI
Intelligent AI platform using LangChain to automatically generate business-relevant data visualizations and insights from raw datasets.
LangChain Streamlit GCP
Machine Learning ↗
LendSafe
FCRA-compliant loan decision explanation system with fine-tuned IBM Granite 350M for transparent, auditable lending decisions.
IBM Granite LoRA Fine-tuning Gradio
Agentic AI ↗
Hamcaller Custom LLM
Custom fine-tuned LLM designed to identify and filter spam calls using advanced NLP techniques to protect users from fraud.
Ollama Fine-tuning NLP
Next.js • Hackathon ↗
TasteBuds Live
Yelp AI API Hackathon submission. Full-stack Next.js 14 + TypeScript app on Vercel — finds your “taste twins” using collaborative filtering and Yelp’s AI API. Python recommendation backend.
Next.js TypeScript Vercel Yelp AI API ▶ Live Site
Full-Stack App ↗
MedicalRAG Live
Production clinical decision-support platform with HIPAA-compliant auth, encrypted sessions, REST API, and RAG backend — deployed on Railway with real users.
FastAPI RAG PostgreSQL Railway
Full-Stack Product ↗
AI-Powered CRM — ContextQA
Built and shipped a React + TypeScript CRM from scratch in 2 weeks — capturing 115+ AI-sourced leads, automating pipeline management, and cutting sales rep processing time by 60%. Delivered 18 qualified prospects at 16% conversion within the first sprint.
React TypeScript Agentic AI REST API
Shipped In
2 Weeks
Leads Captured
115+
Efficiency Gain
60% faster

Credentials

Cloud Certifications

Microsoft Certified: Azure Data Scientist Associate
Credential ID: C2F98ADE9E9ECB9D
Verify ↗
Oracle Cloud Infrastructure 2025 Certified Data Science Professional
Oracle Certification 2025
Verify ↗

Get In Touch

Let's Connect

Open to full-stack AI product engineering roles — building and shipping AI-powered web apps, automations, and internal tools. If you're moving fast and need someone who can go from idea to production, let's talk.

Email
ad2546@rit.edu
LinkedIn
Connect with me
GitHub
View my code
Send a Message