X Recommendation Engine
AI-MLLLMFull-Stack
Chicago, IL — open to work, no sponsorship needed
Shraddha Rao
Full-Stack AI/ML Engineer
Turning "what if..." into working software. Building AI agents, LLM applications, and production systems with GPT-4o, Gemini, Llama, FastAPI, Next.js, and Docker.
Shraddha (Talk to my AI twin)
Tap to start voice chat
Resume
Skills, certifications, and education at a glance.
Skills
Languages
PythonTypeScriptJavaScriptJavaShell / Bash
ML / AI Frameworks
TensorFlow/Kerasscikit-learnXGBoostLightGBMLangChain
Web Frameworks
FastAPIFlaskNext.jsReact NativeStreamlitApollo GraphQL
Algorithms & Techniques
XceptionCNNTransfer LearningThompson SamplingUCBε-GreedySMOTERAG
LLMs & AI APIs
OpenAI GPT-4oGroq Llama 3.3Google Gemini 1.5 FlashHugging FaceDatabricks
DevOps & Cloud
AWS EC2AWS S3AWS IAMAWS LambdaAWS VPCAWS RDSAWS CloudFormationAWS Route 53DockerLinuxVercelRenderHugging Face Hub
Education
Per Scholas Chicago
AWS re/Start — Cloud & IT Fundamentals
Chicago, IL · 2024 – 2025
Flatiron School
Software Engineering Bootcamp
Chicago, IL · 2022 – 2023
Raffles Design Institute
Bachelor of Design in Fashion/Apparel Design
Singapore · 2013 – 2017
Coventry University
Bachelor of Arts (Hons) in Fashion/Apparel Design
Coventry, UK · 2017 – 2018
Work Experience
Where I've shipped things that mattered.
UI Engineer Intern
Patient Studio
Sep 2021 — Mar 2022USA, Remote- Built React Native UI components with Apollo GraphQL queries and mutations for a patient management platform.
- Facilitated daily stand-ups and coordinated cross-functional sprint delivery across design, backend, and QA teams.
- Built dynamic patient form pages with reusable components and REST/GraphQL API integrations, reducing form build time by ~40%.
- Launched a real-time patient–doctor chat widget with WebSocket integration, improving in-app communication response times.
React NativeApollo GraphQLTypeScriptREST APIsWebSocketsFashion Consultant
Streamoid - AI for Fashion Retail
Aug 2018 — Jan 2019India, On-site- Structured high-quality metadata for fashion products to support ML model training and visual search.
- Built a pipeline for inventory categorization and created thousands of activewear image combinations.
- Helped streamline AI-powered fashion tagging systems by annotating product images with bounding boxes.
Data AnnotationMetadata StructuringInventory CategorizationVisual SearchComputer Vision
Projects
End-to-end systems, shipped and documented.
Customer Churn Prediction Model
AI-MLLLMFull-Stack
Brain Tumor Classification
AI-MLLLM
Discord Clone
LLMFull-Stack
AI Coding Agent
LLMFull-Stack
Image Diffusion Web App
LLMFull-Stack
querymori - mcp for databases
LLMFull-Stack
AI Answering System
LLMFull-Stack
Agent Workflow Automation
LLMFull-Stack
Computer Vision App
AI-MLFull-Stack
NLP Text Analysis
AI-MLFull-Stack
AI/ML Research
AI-ML
Blog & Publications
Writing on ML systems, LLMs, and engineering craft.
Medium Articles
View allNo articles loaded — check that the Medium RSS feed is reachable at build time.
Read on MediumPublications
paper2025
Placeholder: Your Publication Title Here
Conference / Journal Name
preprint2024
Placeholder: Another Publication
Workshop / Symposium Name
Talks & Videos
Walkthroughs, talks, and technical deep-dives.
MLFastAPIDocker
Building a Churn Prediction Pipeline — End to End
A walkthrough of building the Bank/Telco churn ensemble system from data ingestion through FastAPI deployment.
Mar 1, 2025
Deep LearningTensorFlowMedical AI
Brain Tumor Classification: Transfer Learning Deep Dive
Hands-on session covering Xception fine-tuning, class balancing, and Streamlit dashboard integration.
Jan 15, 2025
LangChainLLMRecommendation Systems
LangChain Agents + Thompson Sampling for Feed Ranking
Exploring how exploration strategies prevent filter bubbles in the X Recommendation Engine.
Nov 20, 2024
Agent Coding Sessions
Agentic workflows in practice — what I built, what I learned, and what actually saved time.
$ claude_code session
Claude Code
~3 hoursFeb 2025
Building the Churn Prediction Pipeline with Claude Code
// what was built
Complete FastAPI backend + Hugging Face Hub model loader for the Customer Churn Prediction system — from scaffolding to Docker deployment.
key_takeaways:
- →Claude Code handles boilerplate (FastAPI routes, Pydantic models, Dockerfiles) faster than manual coding.
- →Iterative prompting works better than one-shot — break the task into sub-components.
- →Always review generated Dockerfiles for security (non-root user, minimal base image).
$ cursor session
Cursor
~4 hoursJan 2025
Pair-Programming the Recommendation Engine with Cursor
// what was built
LangChain simulation layer with LLM personas and exploration strategy integration (Thompson Sampling, UCB, ε-Greedy) for the X Recommendation Engine.
key_takeaways:
- →Cursor's codebase-wide context is invaluable for refactoring across multiple files simultaneously.
- →GPT-4-level reasoning in the editor catches logical bugs in probability calculations early.
- →Tab completion inside complex lambda chains saves significant time.
$ github_copilot session
GitHub Copilot
~2.5 hoursDec 2024
Fine-Tuning Xception for Brain Tumor Classification with Copilot
// what was built
Two-phase fine-tuning pipeline with custom augmentation layers, class balancing, and evaluation reporting.
key_takeaways:
- →Copilot excels at filling in repetitive TensorFlow/Keras boilerplate (callbacks, metrics, layer configs).
- →Suggestion quality improves dramatically when you write detailed docstrings before the function body.
- →Still need domain knowledge — Copilot suggested incorrect class weight calculation for multi-class imbalance.
Beyond the Keyboard
The things that keep me curious, grounded, and human outside of engineering.
Travel
Exploring new cities and cultures — Singapore shaped how I think about design and density.
Reading
Equal parts technical books (PRML, Designing Data-Intensive Applications) and fiction.
Fitness
Morning workouts keep me sharp — consistency in the gym mirrors consistency in code.
Art & Design
Sketching and digital illustration — the same eye for aesthetics shows up in my UIs.
Volunteering
Chicago community events and tech mentorship for underrepresented folks in STEM.
Accomplishments & Awards
Credentials, competitions, and community contributions.
AWS re/Start Graduate
Amazon Web Services · Mar 2025
Full-time program covering EC2, S3, IAM, VPC, Lambda, CloudFormation, RDS, Route 53, Linux, Python, and Shell scripting.
Generative AI Fundamentals
Databricks · Apr 2025
Covers LLM foundations, RAG pipelines, fine-tuning strategies, and responsible AI practices on the Databricks platform.
Databricks Fundamentals
Databricks · Mar 2025
Core Databricks platform skills: notebooks, clusters, Delta Lake, and data engineering workflows.
Databricks Lakehouse Fundamentals
Databricks · Mar 2025
Lakehouse architecture, Unity Catalog, Delta sharing, and governance best practices.
Python Essentials 1
Cisco Networking Academy · Mar 2025
Python programming fundamentals — data types, control flow, functions, and OOP basics.
Linux Unhatched
Cisco Networking Academy · Dec 2024
Linux command-line fundamentals: file system navigation, permissions, and basic shell scripting.
Developing AI Applications with Python and Flask
Coursera / IBM · Sep 2023
Building and deploying AI-powered web applications using Flask and IBM Watson APIs.
Software Engineering Job Simulation
J.P. Morgan / Forage · Oct 2023
Financial data visualization and JPMorgan's Perspective library in a simulated engineering environment.
X / Twitter
Thoughts on AI, engineering, and building in public. @yourhandle
Shraddha Rao
@yourhandle
🚀 Just shipped the X Recommendation Engine — a full-stack 4-stage ranking pipeline with Thompson Sampling to fight filter bubbles. Built with FastAPI + LangChain + Next.js. Thread on what I learned 🧵
Mar 1, 2025 0 0
Shraddha Rao
@yourhandle
98.9% accuracy on brain tumor MRI classification using Xception transfer learning. Two-phase fine-tuning matters more than people realize. Details in the repo 👇
Jan 10, 2025 0 0
Shraddha Rao
@yourhandle
Hot take: building with Claude Code > writing boilerplate. Used it to scaffold the entire FastAPI + Docker stack for my churn prediction system in an afternoon. The key is breaking tasks into sub-components.
Dec 20, 2024 0 0
Let's Talk
Open to new roles, collaborations, and interesting problems. I respond within 24 hours.