Shubhanshu Shrimali

AI-Native Systems & Game Engineer with 2+ years shipping production games and apps to 300K+ users with Top 10 genre rankings on both App Store & Google Play. Builds autonomous multi-agent systems (LangGraph, MCP, vLLM), self-hosted LLM infrastructure, Graphify code search, Hermes skills/IDE, and high-concurrency multiplayer backends. Manages cross-platform rollout pipelines, app analytics, A/B testing, and ad monetization.

• Grew game audience from zero to 300K+ users, achieving Top 10 genre ranking on both iOS & Android through ASO, A/B tested paywalls, and staged rollouts.
• Architected cyclic multi-agent graphs via LangGraph with persistent state checkpoints, MCP tool servers, and automated eval harnesses using LangSmith with LLM-as-a-judge CI/CD gates.
• Deployed self-hosted vLLM inference on RunPod/Vast.ai with speculative decoding & prompt caching; built trace extraction pipelines for continuous QLoRA/PEFT fine-tuning.
• Shipped production casual, idle, puzzler, and runner titles in Unity (C#) and gamified apps in Flutter/Flame with complete ecosystems (tutorials, daily rewards, leaderboards, IAP, UGUI).
• Managed hybrid monetization: Firebase Analytics, Remote Config A/B tests, ad mediation (AdMob, AppLovin, Unity Ads), IAP funnels, and Crashlytics/Sentry across the 300K+ user base.
• Owned full cross-platform rollout: Web, iOS (Xcode/TestFlight), Android (Play Console), Fastlane automation, GitHub Actions CI/CD, and dynamic color theming pipelines.

• Shipped an MMORPG, 2.5D platformer, and mobile titles; architected authoritative headless servers on AWS via Mirror Networking with lag compensation for 50+ CCU.
• Designed scalable AWS backend (EC2, auto-scaling) for zero-downtime under traffic spikes; reduced memory by 35% and build size by 40% via Addressables.
• Profiled shaders (HLSL) and draw calls via RenderDoc & Unity Profiler, locking 60 FPS on low-tier mobile hardware.

• Built deterministic 2D game loops, state machines, and payout engines in C#; cut memory 25% via object pooling.

• Built XR prototypes with OpenXR & Meta Quest SDK; optimized rendering to sustain 72+ FPS on standalone headsets.

UE5 (C++), GAS, Niagara, Steamworks, AWS EC2

• Open-world RPG with GAS attribute replication, AI behavior trees, vehicle physics, and 64-player dedicated servers.

C/C++, Vulkan, GLSL, ECS, Custom Allocators

• Custom engine from scratch with Vulkan rendering, AI-native subsystems for procedural gen and runtime LLM integration.

Python, Electron, Graphify, MCP, Skills

• Extended Hermes with extra skills, wired Graphify for codebase query and code search, and added the Electron desktop IDE so the agent can search, plan, and edit in one loop.

LangGraph, FastAPI, Pydantic, MongoDB

• Six-agent LangGraph pipeline (research → strategy → copy → QA → forecast → publish) with isolated tools and brand memory.

Python, LangGraph, MCP, vLLM, LoRA/PEFT, Docker

• 24/7 research agents with LangGraph, MCP tools, and trace extraction for literature mining & LoRA dataset distillation.

Next.js, MDX, GenAI Pipelines, DEV.to API

• AI agents that research, draft, fact-check, and auto-publish deep-dive technical articles.