0
Releaf users worldwide
Palo Alto, California · Available for internships & research
Building AI systems for sustainability, accessibility, bioengineering, and astrobiology.
I’m Shengbo “Micheal” Jiang — a high-school junior building multimodal AI models, full-stack applications, and research-driven software systems. iGEM 2024 Gold Medal Tech Lead, USACO Silver, cofounder of Rematter Cooperation, and builder of Releaf, Astrobio-gen, and a locally trained sign-language translation AI.
- iGEM 2024 Gold · Tech Lead
- USACO Silver
- 5,000+ Releaf users
- $2,000+ raised
- Athena AI Labs Intern
Proof, not promises
Numbers from shipped systems, scientific competitions, and production software.
0
Funding raised for Rematter
Gold
iGEM 2024 · Tech Lead, LCG-China
Silver
USACO Competitive Programming
Top 2
National competition prize, China
Full-stack + AI
iOS · Backend · RAG · ML · Web
Selected Work
Systems, not demos.
Each project is treated as a product: architecture, users, constraints, and outcomes. Every system here has shipped, competed, or been published as open source.
Releaf — Ultimate Sustainability
A production iOS app delivering domain-specialized sustainability guidance through AI Q&A and multimodal reasoning.
Astrobio-gen
A multimodal astrobiology inference pipeline for exoplanet habitability and biosignature likelihood mapping.
Sign Language Translation AI
Locally trained ML system translating live speech and video into sign-language output for science education access.
Athena AI Labs — College Advisory
End-to-end AI advisory system: iOS frontend, cloud backend, ChromaDB retrieval, and profile-conditioned RAG recommendations.
iGEM 2024 — Gold Medal, Tech Lead
Computational tools, AMP screening workflows, and software support for a bioactive bacterial-cellulose wound-dressing project.
Sustainability AI Engine
The intelligence layer behind Releaf: domain-focused AI for recycling, upcycling, and environmental reasoning.
About
I build AI systems that survive contact with real users, real data, and real science.
My work sits at the intersection of machine learning, scientific computing, product engineering, and technical leadership. Instead of treating AI as a demo layer, I build complete systems: data pipelines, model logic, backend infrastructure, user-facing interfaces, and measurable product outcomes.
My strongest interests are multimodal modeling, retrieval-augmented generation, graph neural networks, computer vision, accessibility technology, sustainability intelligence, and scientific machine learning. I have applied these across a sustainability AI engine for the Releaf iOS app, a local sign-language translation model for educational accessibility, Astrobio-gen for exoplanet habitability inference, and a college-application advisory system at Athena AI Labs.
As Tech Lead for LCG-China’s iGEM 2024 Gold Medal project, I coordinated computational tools, AMP screening workflows, and software support for bioengineering research. Across my organizations, I often sit at the bridge between product goals, scientific constraints, model design, and implementation.
AI / ML Engineering
Multimodal inputs, RAG pipelines, GNN reasoning, computer vision, and model-evaluation workflows.
Full-Stack Product
User-facing iOS and web apps with backend services, databases, APIs, and product iteration loops.
Scientific Computing
Astrobiology, sustainability science, bioengineering, and educational accessibility tooling.
Technical Leadership
Interdisciplinary student teams, engineering roadmaps, and turning abstract ideas into working systems.
Releaf · Rematter Cooperation · 2024 – Present
Sustainability AI that turns environmental questions into actionable decisions.
Releaf is a production iOS sustainability platform built to reduce the gap between environmental knowledge and real action. Instead of giving generic climate advice, the app answers practical questions — “Can I recycle this?” “How can I reuse this?” “What’s the tradeoff?” — with domain-grounded, mobile-friendly guidance.
- Shipped iOS app built under Rematter Cooperation with advanced AI features fine-tuned for sustainability Q&A.
- 5,000+ users worldwide with app instrumentation and feedback loops driving iteration.
- $2,000+ raised in project funding and a top-two prize in a national competition in China.
- My scope spanned AI flows, mobile/backend engineering, and the product roadmap across the full system.
5,000+users worldwide
Top 2China competition
Architecture — five layers, one product
01iOS Client
Question input · assistant UI · educational cards · feedback
02Backend API
Routing · session · AI inference · logging & analytics
03Sustainability AI
Domain prompting · RAG pipeline · safety & quality filters
04Knowledge Layer
Recycling rules · materials · upcycling strategies · local orgs
05Evaluation & Iteration
Feedback · response quality · safety · instrumentation
Sustainability AI Engine
A domain-focused AI system for recycling, upcycling, and environmental reasoning.
The intelligence layer behind Releaf. It converts messy user questions into structured queries, retrieves sustainability knowledge, reasons over constraints, and returns practical guidance — quick answer, why it matters, what to do next, cautions, better alternatives, and local action when available.
Input Understanding
Intent classification, material detection, query normalization across nine sustainability intents.
Retrieval Layer
RAG over recycling rules, material properties, impact explanations, reuse strategies, and local organizations.
AI Reasoning
Synthesizes user intent, object context, retrieved evidence, safety, and clear action steps.
Response Formatting
Mobile-friendly blocks so the answer is usable on the go, not a wall of text.
Phase 1
Text sustainability assistant · mobile integration · prompt/response control.
Phase 2
Multimodal waste recognition · image upload · localized recycling suggestions.
Phase 3
Environmental action network · charity & community matching · impact tracking.
Sign Language Translation AI · Science Seed Project
Local multimodal AI for more accessible communication.
A locally trained, accessibility-focused ML system exploring live translation from speech and video into sign-language output — designed to help students with hearing-related barriers access science education.
Useful sign-language translation is not letter spelling. It requires sentence-level meaning, grammar, timing, and continuous motion. I designed this system as a multimodal challenge combining audio processing, language modeling, computer vision, sequence modeling, and interface design — trained and debugged on local hardware rather than behind a black-box API.
01
Not letter spelling
Beyond alphabet classification — sentence, grammar, and context handling.
02
Real-time constraints
Live audio/video processing at latency low enough to be usable in a classroom.
03
Dataset limits
Uneven coverage, low sample counts, and generalization across unseen vocabulary.
04
Domain vocabulary
Science terms rarely appear in common sign corpora — fallback strategies required.
Input
Microphone · video stream · educational content
Preprocess
Speech-to-text · landmark detection · temporal segmentation
Understanding
Sentence parsing · concept mapping · educational vocabulary
Sign Plan
Vocabulary lookup · gesture sequence · temporal smoothing
Output
Visual sign display · captions · learning interface
Student Profile
Academic interestsAI · Bioengineering
Target fieldComputer Science
Activity signalResearch · iGEM · iOS
Retrieval
ChromaDB · 18 sources
Match Score
0.92
Profile-conditioned
Recommended Programs
RSISummer research
Stanford AIProgram · fit 0.94
MIT EECSMajor alignment
Athena AI Labs · Software Intern · 2025 – Present
From static admissions data to profile-aware recommendations.
An end-to-end AI advisory platform combining an iOS frontend, cloud-hosted backend, and a RAG pipeline backed by ChromaDB. The system converts academic profiles into structured retrieval queries and returns university, major, and summer-program recommendations with explanations of fit.
- Built components of the advisory system across frontend, backend, and retrieval layers.
- Used ChromaDB as the vector database for embedded admissions and program data.
- Designed RAG workflows for profile-conditioned recommendations and explainable outputs.
- Contributed to an end-to-end architecture, not isolated prototype code.
A useful advisory system depends on data quality, retrieval precision, backend reliability, frontend clarity, and recommendations users can trust — not just model access.
iGEM 2024 · LCG-China · Gold Medal · Tech Lead
Technical leadership across computational biology, software, and engineered biomaterials.
I led technical development for a synthetic-biology project combining antimicrobial peptides with a bioactive bacterial-cellulose dressing concept — a modular system designed to present antimicrobial activity on contact with moisture, ions, or wound exudate.
My responsibility was turning biological design requirements into computational workflows. That meant building sequence design and evaluation tools, integrating docking and structure-prediction outputs, engineering AMP-screening data pipelines, and authoring technical documentation — all on a fixed competition timeline, across interdisciplinary subteams.
01
Biological Goal
Bacterial-cellulose dressing functionalized with modular antimicrobial peptides.
02
Candidate Collection
AMP sequences, literature references, annotations, and activity indicators.
03
Computational Screening
Sequence property analysis, candidate ranking, docking & structure integration.
04
Wet-Lab Support
Selection pipeline, experimental planning, tracking, and engineering feedback.
05
Software & Docs
Internal tools, visualization, reporting, and competition deliverables.
Habitability0.71
Biosignature0.34
Uncertainty±0.12
Astrobio-gen · Sole Developer · Open Source
Multimodal astrobiology AI for exoplanet habitability inference.
An open-source pipeline that fuses stellar activity, orbital and planetary parameters, transmission and emission spectra, and a Milky-Way-scale galactic prior into a unified representation — producing habitability scores and biosignature likelihood with uncertainty estimates rather than binary claims.
- Preprocessing for schema normalization, missing-value control, physical range validation, and modality tensor construction.
- Advanced attention, hierarchy learning, and noise-reduction methods for noisy scientific data.
- Trained and evaluated on public catalogs such as the NASA Exoplanet Archive.
- Uncertainty-aware prediction — no overconfident habitability claims.
- Targeted for an ISEF research submission as an open-source research platform.
Stellar activity
Orbital params
Planetary features
Spectra
Galactic prior
CarbonTrack · Tech Lead · 2023 – 2025
A student-built carbon-footprint tracking and education platform.
Built databases and web applications powering the Carbon Footprint Club’s tracking platform at Hamden Hall. Deployed a public site, organized student contributors, and maintained code quality as the project grew into a community climate-education tool.
Science Seed Project · Founder · 2024 – Present
Free online science education for children with limited resources.
Founded a free science-learning platform connecting my interests in science, AI, and social impact. Home to the sign-language translation AI — accessibility is part of the same mission: help more students reach scientific knowledge.
Experience & Leadership
From student clubs to shipped AI products.
A compact record of the roles where I’ve led engineering, coordinated teams, or built the software that ships.
-
Software Intern
Building a full-stack AI college-application advisory platform with iOS frontend, cloud backend, ChromaDB vector retrieval, and RAG-based profile-conditioned recommendation workflows.
- Full-stack AI system design
- iOS + cloud integration
- ChromaDB retrieval
- University, major, and summer-program recommendations
-
Cofounder & Tech Engineer
Cofounded and engineered Releaf, a production iOS sustainability app with domain-specialized AI features, backend support, and mobile product execution.
- Shipped Releaf iOS app
- 5,000+ users worldwide
- $2,000+ funding raised
- Top-two prize · China competition
-
Tech Lead — iGEM 2024 Gold Medal
Led computational tools, software workflows, and AMP screening support for a bioactive bacterial-cellulose wound-dressing project.
- AMP sequence design/evaluation tools
- Docking & structure integration
- Technical documentation
- Interdisciplinary coordination
-
Founder
Founded a free science-education platform and developed locally trained sign-language translation AI to support accessible learning.
-
Tech Lead
Built databases and web applications for a student carbon-footprint tracking platform and organized student technical contributors.
-
Team Leader
Led training and contest strategy; achieved highest score in the Greater New Haven League in 2024.
Awards & Honors
Recognized work, from biology labs to algorithmic contests.
Gold
iGEM 2024 Gold Medal
Tech Lead · LCG-China
Awarded for a synthetic-biology project on antimicrobial-peptide-functionalized bacterial cellulose and the supporting computational/software workflows.
Silver
USACO Silver — 2024
Competitive Programming
Achieved USACO Silver, demonstrating algorithmic problem-solving ability and competitive programming strength.
1st
Greater New Haven League
Highest Scorer · 2024
Highest scorer in the Greater New Haven League as Math Team leader at Hamden Hall.
Top 2
National Competition, China
Rematter Cooperation / Releaf
Top-two national prize for Rematter Cooperation and $2,000+ in project funding raised through the program.
Technical Skills
Grouped by how I actually use them.
01
AI / Machine Learning
- Multimodal modeling
- Retrieval-augmented generation
- Graph neural networks
- Computer vision
- Attention mechanisms
- Hierarchy learning
- Noise-reduction techniques
- Parameter-efficient training
- Local model training & evaluation
- Uncertainty-aware modeling
- Scientific machine learning
02
Software Engineering
- Python 3
- iOS app development
- Web app development
- Backend APIs
- RESTful API design
- Database design
- Cloud-hosted backends
- Git / GitHub
- Full-stack architecture
03
Data / Scientific Computing
- Data preprocessing
- Schema normalization
- Missing-value handling
- Physical range validation
- Vector databases
- ChromaDB
- Scientific dataset integration
- Model evaluation workflows
04
Product & Leadership
- Technical roadmap planning
- Cross-functional coordination
- Student contributor mentoring
- Product iteration
- Technical documentation
- Competition delivery
- Research-to-product translation
Frequently Asked
A few questions I get from recruiters, professors, and collaborators.
Short answers — follow any of the project sections above for full technical depth.
What kind of AI systems do you build?
Multimodal systems that have to survive contact with real users and real data. My work spans sustainability Q&A, accessibility translation, exoplanet habitability inference, bioengineering screening tools, and retrieval-augmented recommendation platforms.
What’s the strongest project on this portfolio?
Releaf is the most visible — a shipped iOS product with 5,000+ users and a funded team. Technically, the Sustainability AI Engine and Astrobio-gen represent my deepest system design and scientific ML work.
What is Releaf and what makes the Sustainability AI model different?
Releaf is a production iOS sustainability platform. The Sustainability AI Engine behind it uses retrieval-augmented generation, intent classification, and response formatting tuned for practical action — not generic environmental slogans.
What is the Sign Language Translation project trying to solve?
Accessible science education. The system explores live speech- and video-to-sign translation beyond alphabet classification — sentence-level meaning, timing, and domain vocabulary. It is under active development, trained locally.
What did you do as iGEM 2024 Tech Lead?
Led computational tools, AMP screening workflows, and software support for LCG-China’s Gold Medal bioactive bacterial-cellulose wound-dressing project, coordinating interdisciplinary subteams on a fixed timeline.
What are you building at Athena AI Labs?
An end-to-end AI college-application advisory system: iOS frontend, cloud backend, ChromaDB retrieval, and RAG-based profile-conditioned recommendations for universities, majors, and summer programs.
Is Astrobio-gen open source?
Yes. It is a multimodal astrobiology inference pipeline being developed openly, with uncertainty-aware predictions and a roadmap toward an ISEF research submission.
Are you available for internships, research, or collaborations?
Yes. I am open to research collaborations, AI engineering internships, scientific ML opportunities, sustainability technology projects, and technical partnerships. Reach me at shengbojmicheal8866@gmail.com.
Let’s work
My projects share one pattern: AI systems built around real constraints.
Sustainability AI has to be practical enough for mobile users. Sign-language translation has to respect real-time accessibility. Astrobiology modeling has to handle incomplete scientific data and uncertainty. Bioengineering software has to support wet-lab decisions, not just look impressive in a demo. That’s the engineering I want to keep doing — technically ambitious systems that survive contact with real users, real data, and real scientific problems.