Live
CareerLoom
ProductivityHelps job seekers turn experience, resume notes, and interview prep into material they can actually use.
Go / realtime / backend infrastructure
Go realtime backendwith replaceablecommunication layers
I have built Go backends and realtime infrastructure, turning WebRTC, IM, and product events into clear contracts. Services can deploy independently, and clients do not have to be locked to one communication path.
35
On the App Store
Apps keep evolving; the sections below break the background into specific capability lenses.
Go / realtime / backend infrastructure
Go realtime backend with replaceable communication layers
I have built Go backends and realtime infrastructure, turning WebRTC, IM, and product events into clear contracts. Services can deploy independently, and clients do not have to be locked to one communication path.
Go · gRPC · PostgreSQL · realtime
Backend skill shows up when real traffic arrives
I use Go for high-concurrency APIs, realtime messaging, and infrastructure glue. Layers, raw SQL, caches, and message paths all need clear ownership when something breaks.
I care about contracts and deployment shape: services can ship independently, dependencies stay isolated, and the product is not locked to one transport.
- Go / Gin REST APIs
- gRPC publish and WebSocket subscriptions
- PostgreSQL / Redis / raw SQL
- WebRTC / IM plugin abstraction
- Independently deployable realtime infrastructure
Positioning
Product work, with systems depth
The best fit is not a single screen or endpoint. It is work that needs mobile, backend, AI, release, and operations connected. Earlier low-level and fintech work still shapes how I think about boundaries, stability, and long-term maintenance.
Portfolio
Independent apps
Live
CreatorKit
CreativeWriting and editing support for creators who need to move from idea to publishable draft faster.
Live
Lullacraft
HealthA bedtime app focused on rhythm, sound, and a low-friction wind-down experience.
Live
Memory Copilot
ProductivityA long-term memory assistant that turns scattered notes into material worth revisiting.
Live
SpeakLab
EducationConversation practice with repeatable drills and weakness review instead of generic feedback.
Live
Spritewood
CreativeGenerates bead-art patterns and color plans so a visual idea can become a makeable bead design.
Core capabilities
Turn complex work into a shippable chain
Mobile systems and native feel
iOS, React Native / Expo, and Flutter; when native work, startup flow, or performance detail is needed, I do not stop at the business layer.
Backend and cloud-native
Go / Gin, PostgreSQL, Redis, gRPC, Docker, AWS EKS / Kubernetes, from API contracts through deployment and rollback.
AI products and agent orchestration systems
LLMs, RAG, long-term memory, prompts, and agent orchestration, review loops, and release checks for AI-assisted delivery.
CI/CD and operations
Builds, releases, health checks, alerts, certificates, and domains are debugged together. Releases need evidence, and failures need a layer.
Complex systems
Beyond independent apps, production systems too
Public technical signals highlighted here: realtime communication, Go backend, cloud-native migration, CI/CD, ops checks, and agent orchestration.
Realtime communication and Go backend
IM, WebRTC, product events, and client subscriptions split behind clear contracts; backend services can deploy independently, and the client is not locked to one communication layer.
Go · gRPC · WebSocket · PostgreSQLEKS migration and release governance
Production migration planning for AWS EKS, including container images, Helm, Ingress, secrets, database, cache, maintenance window, rollback, and health checks.
AWS EKS · Kubernetes · Helm · CI/CDBest-fit workMobile / full-stack AI productsGo backend / cloud-native migrationAgent orchestration / ops automation
Open-source work
Open-source projects
transcendence-memory
A self-hosted memory service for AI agents, covering retrieval, RAG, and multimodal ingestion.
- Container-level isolation with cross-container retrieval when needed
- LanceDB vector search paired with LightRAG knowledge-graph retrieval
- Multimodal ingestion for PDFs, images, and tables
- Connection-token setup for Claude, Gemini, Cursor, and similar tools
MITLanceDBLightRAGRAG-AnythingPython
View GitHubStack
A fuller full-stack map
React Native / Expo mobile delivery
Swift / SwiftUI native iOS
Flutter / Dart cross-platform
Next.js / React 19 frontend
Go / Gin · gRPC · realtime paths
PostgreSQL · Redis · raw SQL
LLM gateway · RAG · long-term memory
Subscriptions · StoreKit · store compliance
Payments / wallet / global payout experience
Cloudflare Pages · Nginx · edge caching
AWS EKS · Kubernetes · Helm
Multi-agent orchestration · collaboration boundaries · result review
Path
From engineering front lines to independent delivery
1
About 13 years
Internet and fintech · iOS / low-level / payments
Objective-C / Swift, audio/video, OpenGL, high-performance SDKs, wallet and payment modules, plus small-team technical leadership.
2
Full-stack stage
Global social / AI product · Go backend and Flutter client
Go / Gin, PostgreSQL, Redis, gRPC, realtime chat, wallet transactions, AI chat, and cross-platform client work.
3
Now
Sanva · independent AI apps and engineering foundation
Shipping independent apps while building scaffolds, LLM gateway, RAG memory, CI/CD, ops checks, and agent-assisted delivery systems.
Design and implementation
Consistency and product work
The visual system is mine too: brand color, interface, and components. The apps follow Apple HIG, but each keeps its own character.
Indigo
Ember
Teal
Cream
Peach
Lavender
Each app has its own color, icon, and palette, while sharing a common design language.
How I work
From idea to release
01
Validate the problem first
Two weeks is enough to get to a working prototype. Real feedback decides whether it continues.
02
Bring in AI after that
AI tools help move faster, but the code is still reviewed by hand.
03
Verify at the end
Before submission I run the full features.yaml checklist: no fake data, no production mocks.
04
Keep closing the loop after release
After release, I revise quickly from real user feedback.