Yupeng is available for hire
Hire YupengYupeng Gu
Verified Expert in Engineering
Full-stack Developer
Nanjing, Jiangsu, China
Toptal member since May 17, 2021
Bio
Yupeng is a senior AI and full-stack architect building production-ready systems from 0 to 1. Spanning React Native, Python, and AWS Serverless backends, he specializes in cost-effective LLM workflows and low-latency architectures. Having built platforms single-handedly, he brings battle-tested pragmatism to tech choices—avoiding over-engineering for rapid iteration and clean execution.
Portfolio
Echo Project (AI Social & Gaming Platform)
Python 3, FastAPI, Google Cloud Platform (GCP), Cloud Run, Firebase...
Pepster GO (Location-based IoT & AR Platform)
Unity, Augmented Reality (AR), Near-field Communication (NFC)...
Pepster Limited
React Native, Expo, React, Figma, JavaScript
Experience
- Python 3 - 5 years
- Expo - 4 years
- Python - 3 years
- Artificial Intelligence (AI) - 3 years
- JavaScript - 2 years
- LLM Integration - 2 years
- FastAPI - 2 years
- React - 1 year
Preferred Environment
MacOS, Linux, Ubuntu, Gemini, Expo, React Native
The most amazing...
...professional project I've worked on was the Echo Project.
Work Experience
Lead AI & Full-stack Architect
2025 - PRESENT
Echo Project (AI Social & Gaming Platform)
- Engineered a scalable AI-driven social gaming engine from 0 to 1 as the sole full-stack architect, leveraging Python (FastAPI), React Native, and Vertex AI to power real-time, multi-agent text simulations.
- Achieved near-zero fixed infrastructure cost at launch by architecting a purely serverless, pay-as-you-go back end, completely eliminating heavy monthly cloud overhead while maintaining instant multi-tenant scaling.
- Eliminated user-perceived LLM inference latency (sub-second response) using async GCP Cloud Tasks message queues, guaranteeing seamless state machines and session synchronization for high-concurrency users.
- Solved long-context amnesia in AI agents by building custom knowledge-graph modeling pipelines, significantly boosting long-term user engagement and story continuity over extended sessions.
- Secured user data privacy and compliance by implementing GCP KMS envelope encryption for real-time memory cipher packets, laying an enterprise-grade zero-trust security foundation.
Technologies: Python 3, FastAPI, Google Cloud Platform (GCP), Cloud Run, Firebase, Cloud Firestore, React Native, Expo, Neo4j, React, JavaScript, AI Agents, Agentic AI, Prompt Engineering, Gemini, Artificial Intelligence (AI), OpenAI API, Python, Large Language Models (LLMs), LLM Integration, Knowledge Graphs, GraphRAG
Lead R&D & IoT Solutions Architect
2025 - 2025
Pepster GO (Location-based IoT & AR Platform)
- Architected a location-based IoT and AR platform from R&D to MVP, leveraging Unity 3D, Niantic Maps, and AWS Serverless to power localized matchmaking and real-time device states.
- Achieved near-zero monthly fixed cloud overhead by building a 100% event-driven, pay-as-you-go back end with AWS SAM, DynamoDB, SQS, and Cognito—delivering instant scalability without expensive idle server costs.
- Pivoted hardware strategy to eliminate real-world tracking failures, replacing light-sensitive Vuforia VuMark AR with a custom, highly reliable NFC proximity protocol that drastically improved field success rates.
- Accelerated R&D and prototyping cycles by leveraging modern AI-assisted engineering workflows, enabling rapid technical feasibility audits and swift architecture validation under tight startup timelines.
- Mitigated safety and compliance risks by establishing deep-dive location-data auditing frameworks, balancing user safety with cost-efficient deployment strategies.
Technologies: Unity, Augmented Reality (AR), Near-field Communication (NFC), Amazon Web Services (AWS), AWS Serverless Application Model (SAM), AWS Lambda, Amazon DynamoDB, Amazon Cognito, C#, Python 3, Gemini, Artificial Intelligence (AI), Generative Design, Python, LLM Integration, FastAPI
Lead Full-stack Architect
2023 - 2025
Pepster Limited
- Rebuilt the flagship cross-platform app from scratch using React Native and Expo, replacing legacy code bases with a unified single-source architecture that cut deployment cycles and maintenance overhead in half.
- Stepped in as the primary technical anchor to stabilize a legacy Node.js/relational back end, eliminating frequent microservice regressions, patch-fixing architectural bottlenecks, and drastically lowering server downtime.
- Contrasted high-maintenance, server-bound legacy setups with modern, cost-efficient patterns, enforcing strict API contracts and modular DB boundaries to safeguard reliability under aggressive launch deadlines.
- Bridge-engineered Figma designs into deterministic React components, establishing seamless design-to-code workflows with international design teams to guarantee pixel-perfect, high-performance UI micro-interactions.
- Mentored and led a distributed engineering team (1–5 devs), implementing rigorous code reviews and modern CI/CD pipelines to elevate code quality and execution speed across the engineering lifecycle.
Technologies: React Native, Expo, React, Figma, JavaScript
Full-stack and Mobile Software Engineer
2021 - 2021
Pepster
- Communicated with the client about the initial idea and clarified the logic.
- Designed the workflow of the Pepster app and the simple, functional UI on Figma from scratch.
- Created and developed the back end using the Django REST Framework from scratch.
- Deployed the Django REST back-end with Kubernetes on Amazon EKS.
- Designed and developed the iOS app from scratch using SwiftUI.
- Published the app among the internal testing users with TestFlight.
Technologies: SwiftUI, Swift 5, iOS, Amazon Web Services (AWS), Python 3, Django, Django REST Framework, Django ORM, Celery, Kubernetes, Amazon EKS, Figma, Model View ViewModel (MVVM), Xcode, Full-stack, Full-stack Development, JavaScript, SQL, MySQL, GHUnit, Python
iOS Mobile Developer
2021 - 2021
Abode Systems LLC
- Rewrote the Koko Face Yoga app with SwiftUI based on the legacy Flutter version.
- Created and added the new module of face symmetry analysis using Swift and Google ML Kit.
- Beta-tested the new version of the app in TestFlight and published the latest version to the App Store.
- Modified the email subscription functionalities using the Mailchimp API and the Firebase Cloud Functions with Node.js.
Technologies: Swift 5, SwiftUI, Firebase iOS SDK, ML Kit, Node.js, Firebase Cloud Functions, Model View ViewModel (MVVM), Xcode, Firebase
iOS Software Engineer
2016 - 2020
Apponda (Lovefoodies Company)
- Built the iOS application from scratch with Swift, including the UI with Auto Layout and HTTP network interface to connect to the RESTful API back end. Integrated with several third-party libraries such as Google Maps, Stripe, and Firebase Messaging.
- Implemented unit and UI testings of the iOS application with XCTest and the CI/CD process of the iOS application using Docker and Jenkins.
- Designed and experimented with GAN networks that generate fake icons automatically with Python, TensorFlow, Keras, and cloud GPU instances.
Technologies: iOS, Xcode, Swift, UIKit, SwiftUI, Firebase iOS SDK, XCTest, Docker, Jenkins, TensorFlow, Keras, HTML, CSS, JavaScript, SQL, MySQL, Firebase, Agile
Research Assistant
2018 - 2019
International Technological University
- Implemented indoor robot mapping and navigation under the ROS system.
- Improved robot obstacle avoidance functions by integrating object detection neural networks with the navigation stack using C++.
- Built a campus robot guidance system by integrating a voice recognition framework with the navigation stack using Python and C++.
- Created a simple app in Swift to remotely visualize the map and control the Turtlebot using iOS devices.
- Generated virtual 3D worlds of robot simulation using procedural content generation methods and experimented with Turtlebot auto exploration under the generated virtual worlds using deep reinforcement learning with TensorFlow and Keras.
Technologies: Robot Operating System (ROS), Simultaneous Localization & Mapping (SLAM), Python 3, C++, Swift, TensorFlow, Keras, Procedural Generation, Agile, Python
Experience
Sphere Fluid Effects (Metal)
https://github.com/freegyp/noisySphereApplied Gabor noise algorithm in Metal and Swift to create an animation of procedural noise effects like dynamic fluid on a sphere. Also implemented in a later version glowing effects using image processing techniques.
256x256 Face Image Generation (StyleGAN)
Implemented the model with TensorFlow Keras based on the paper of StyelGAN and trained the model on Google Cloud AI Notebook with one Nvidia Tesla v100 GPU. We used the dataset CelebA-HQ and chose the image resolution to be 256x256.
GypChat
A simple chat application implemented with Swift and SwiftUI with MVVM pattern. Used SDK of Firebase, including Firestore, cloud functions, and authentication for the back-end functionalities. Also added unit test cases (XCTest).
Note: Some functionalities might not run properly because the Firebase project was deleted after the project was finished.
Note: Some functionalities might not run properly because the Firebase project was deleted after the project was finished.
Toptal React Accelerator
The Toptal React Accelerator is an exclusive learning program that educates JavaScript developers in React, Next.js, Redux, and GraphQL. This is a 2-month program that consists of rigorous learning exercises and hands-on activities along with a stringent evaluation process, comprehensive learning assessments, and an Invoice Management System. Within this invoicing application, users are able to track invoices, clients, and company information.
Education
2018 - 2020
Master of Science Degree in Electrical and Computer Engineering
International Technological University - San Jose, CA, USA
2013 - 2015
Master of Science Degree in Computer Science
College of William and Mary - Williamsburg, VA, USA
Certifications
OCTOBER 2022 - PRESENT
Toptal React Accelerator Graduate
Toptal, LLC
MAY 2020 - PRESENT
Google Code Jam Round 2 Qualifier
MAY 2018 - PRESENT
World CodeSprint 13 Silver Medal
HackerRank
JULY 2017 - PRESENT
Deep Learning Nanodegree
Udacity
Skills
Libraries/APIs
Firebase iOS SDK, Keras, Node.js, React, TensorFlow, NumPy, Django ORM, MobX, OpenAI API
Tools
Amazon EKS, Atom, Xcode, Jenkins, Celery, Figma, Expo, Amazon Cognito, GraphRAG
Languages
Swift, Python, Python 3, Swift 5, C++, HTML, CSS, JavaScript, SQL, TypeScript, GraphQL, C#
Frameworks
SwiftUI, UIKit, Metal, Django, Django REST Framework, Redux, XCTest, Next.js, Material UI, GHUnit, React Native, Unity, AWS Serverless Application Model (SAM)
Paradigms
Model View ViewModel (MVVM), Agile
Platforms
iOS, Linux, Ubuntu, Firebase, Docker, Amazon Web Services (AWS), Kubernetes, MacOS, AWS Lambda, Google Cloud Platform (GCP), Cloud Run
Storage
MySQL, Amazon DynamoDB, Cloud Firestore, Neo4j
Other
Algorithms, Computer Graphics, FastAPI, Large Language Models (LLMs), LLM Integration, Machine Learning, Robot Operating System (ROS), Simultaneous Localization & Mapping (SLAM), Procedural Generation, Shaders, Deep Neural Networks (DNNs), Firebase Cloud Functions, ML Kit, Full-stack, Full-stack Development, Augmented Reality (AR), Near-field Communication (NFC), AI Agents, Agentic AI, Prompt Engineering, Gemini, Artificial Intelligence (AI), Generative Design, Knowledge Graphs
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring