Public resume
Sanva
Full-stack Product Engineer
13 years across internet and fintech engineering, from iOS, audio/video, OpenGL, high-performance SDKs, and wallet/payment systems into full-stack architecture.
Now building independent AI apps while maintaining a delivery pipeline across scaffolds, LLM gateway, RAG memory, CI/CD, ops checks, and App Store release.
This lens focuses on realtime systems, data boundaries, and service design.
Overview
I have 13 years of engineering experience across internet and fintech products. The early core was iOS and lower-level work: Objective-C / Swift, audio/video codecs, OpenGL ES + C++ rendering, high-performance SDKs, wallet, and payment modules. What stayed with me was not a title, but a working standard: clear layers, stable contracts, and no postponing performance or compliance until after launch.
The work later expanded into full-stack and platform engineering. I have built Go / Gin backends, PostgreSQL, Redis, gRPC, realtime communication, Flutter clients, and production migration plans for AWS EKS covering Helm, Ingress, secrets, databases, caches, rollback, and health checks. Now I build AI apps independently, so product, design, mobile, backend, AI, subscriptions, CI/CD, ops, and release all sit on my desk.
Backend work trained me in concrete habits: define contracts first, keep the data layer understandable, and make realtime paths debuggable. In AI products, I use the same habit to keep the model, client, and server replaceable and maintainable.
Experience
From mobile systems to full-stack AI products and platform engineering
About 13 years · internet and fintech front lines
Internet / fintech · iOS and low-level engineering
Long-running work in mobile and low-level systems: audio/video, high-performance SDKs, wallet/payment modules, cross-platform integration, and small-team technical delivery.
- Low-level iOSDeep Objective-C / Swift iOS engineering: large-app modularization, memory and startup performance, stability hardening; also low-level high-performance SDK work that packaged reusable capabilities for multiple teams and products.
- A/V & renderingLow-level audio/video codec SDK work: encode/decode pipelines, player cores, playback at tens-of-millions scale; OpenGL ES + C++ high-performance graphics rendering pipelines, pushing frame rate, power, and quality to ship-ready bars on mobile.
- Fintech / cross-platformCore fintech modules: wallet / payment SDKs, cross-platform integration, and subscription-related flows under high stability, compliance, and security requirements; CI/CD engineering connected native capabilities with cross-platform delivery.
- Team / methodLed small teams of two to five on technical planning and efficiency work. The lasting value was the habits that still stay with me: layer first, align contracts first, keep concurrency and state under control, and verify thoroughly before release.
Startup · full-stack & architecture
Global social / AI product · Full-stack Engineer
Owned core modules across a high-concurrency backend and cross-platform client for an overseas product, moving my work from client engineering into full-stack architecture.
- BackendHigh-concurrency REST APIs in Go + Gin with strict layers (handler / service / db), raw SQL without ORM, PostgreSQL / Redis / object storage / messaging, covering realtime messaging, wallet transactions, identity verification, and AI chat.
- RealtimeOwned the realtime chat path: backend gRPC publish → realtime gateway → client WebSocket subscribe, with channel namespaces by domain; led a full code audit for a major version migration of the realtime service.
- ClientCross-platform client in Flutter / Dart: third-party login, multi-step identity verification, user profile, realtime chat UI, content feed.
- DisciplineThis stage made end-to-end collaboration more important: clarify API contracts, data states, and delivery boundaries first, then move client and backend work forward with less rework.
- StackGo / Gin · PostgreSQL · Redis · gRPC · realtime gateway · Flutter / Dart · Docker · cloud infrastructure.
Present · independent
Sanva · independent AI app portfolio
Now I continue to build independent AI apps. More importantly, I maintain a reusable delivery system: scaffolds, mobile foundation, AI infrastructure, release flows, ops checks, and agent orchestration systems.
- PipelineEvery new app starts from a shared scaffold with the stack, theme, subscriptions, store assets, and quality checks already wired. Research, design, build, acceptance, and release are made as repeatable as possible, with lessons flowing back into the template.
- Prompt eng.Prompts are not one-off copy. They are recipe sources: inputs, constraints, counterexamples, output shape, and checks are kept separate so human judgment survives model or task changes.
- OrchestrationFor multi-agent work, I own decomposition, dispatch, review, and close-out; work units have clear ownership and result standards. Clear ownership, isolated branches, and script checks keep parallel work from becoming chaos.
- AI infrastructureI build and maintain the AI engineering base: multi-model LLM gateway, RAG / long-term memory, conversational product experience, AI data consent, automated checks, and production checks. These are the systems public products keep using.
- StackExpo / React Native / TypeScript (strict) · Swift / SwiftUI · Go / Gin · PostgreSQL · Redis · gRPC · Docker · AWS EKS / Kubernetes · Helm · CI/CD · LLM gateway · RAG / long-term memory · agent-assisted delivery systems.
Stack
Stack grouped by delivery path
Mobile
- Native iOS (Objective-C / Swift / SwiftUI)
- React Native · Expo · TypeScript (strict)
- Flutter / Dart cross-platform clients
- Launch flow · onboarding state machine · native plugins and performance issues
Backend / Architecture
- Go / Gin (high-concurrency · layered)
- PostgreSQL · Redis · raw SQL · gRPC · WebSocket
- Realtime communication / wallet transaction / AI chat backend modules
- Modular system design · frontend-backend contracts · replaceable infrastructure
AI / LLM
- Multi-model LLM gateway · weighted routing · fallback
- RAG retrieval · long-term memory · container isolation
- Conversational product experience · AI consent and store compliance
- Prompt recipe sources · output checks · negative constraints
Agent Orchestration / Automation
- Agent orchestration and review loops
- isolated branches and ownership maps
- Browser-based acceptance checks
- Documentation, scripts, and long-term memory feedback loops
DevOps / Cloud-native
- AWS EKS · Kubernetes · Helm · ALB Ingress
- Docker / Compose · Nginx · PM2 · Cloudflare Pages
- CI/CD (validate / OTA / build / submit) · three-layer health checks
- Certificates / domains / alerts / ops checks / production cutover runbooks
Commerce / Store Release
- In-app purchase · subscriptions · StoreKit 2
- App Store Connect API · build and submission flow
- Store assets · ASO copy · localization
- AI compliance copy · privacy wording · account deletion flow
Public work
Released apps
Selected apps live on the App Store, grouped by category — tap to open the public download page.
Creative
Productivity
Health
Education
Lifestyle
AI resume Q&A
Ask about projects, technical range, and resume details.
Use it for a quick read on Sanva’s mobile, backend, AI product, cloud deployment, CI/CD, and App Store delivery experience.
How I work
How I move work forward
Break the delivery chain first
Split goals into product, client, backend, AI, release, and ops layers, each with inputs, outputs, and checks.
Put complexity inside boundaries
For low-level performance, payments, realtime systems, or EKS migration, start with state, contracts, rollback, and observability.
AI executes; it does not own judgment
Prompts, RAG, agent orchestration, and scripts serve human judgment; parallel work needs scoped boundaries before and review after.
Stay responsible after release
CI/CD, health checks, alerts, certificates, and domains are part of delivery. Launch is not the end.
If you need someone who can own the full chain
Mobile, Go backend, AI products, EKS migration, CI/CD, ops checks, or agent orchestration systems can all start from a concrete problem.
