Harsh Rastogi · AI Product Engineer at Modelia
I own generative-AI systems end to end: the product decision, the model in the loop, the billing underneath and the agent that ships it. Currently leading AI product at Modelia.
Driven by product and obsessed with detail, I take AI ideas from a sketch to systems people rely on, and set the product direction they serve.
At Modelia I lead AI product for a generative image & video platform: the AI video pipeline, the metering engine under every generation, and the agents that support customers. Before that I built AI hiring automation at Asynq, a university LMS for 5K+ learners at CCS University, and defence infrastructure at Bharat Electronics Limited.
Production generative and agentic AI, from image and video pipelines to autonomous agents with LLM integration and evaluation.
Roadmaps, prioritization and revenue-driving features, set by someone who also ships the code behind them.
Production APIs, microservices and modern frontends with TypeScript, Node.js, React, Remix, tRPC, Prisma and PostgreSQL.
AI engines, merchant-facing apps, creative tools and internal systems, from architecture to production on Azure and AWS.
Harsh Rastogi is an AI Product Engineer at Modelia, a generative AI image and video platform for fashion e-commerce. He builds generative-AI products and multi-agent systems end to end, co-founded the multi-agent platform SelfAgentic, and created carcode, an open-source way to code from your car with Claude Code.
An AI Product Engineer owns an AI feature from the product decision to production: choosing where a model belongs, building the pipeline and the guardrails around it, metering and billing usage, evaluating quality, and shipping it to real users. Harsh does this across Modelia's studio, billing engine, support agents and Shopify app.
Modelia's generative image and video platform and its usage-billing engine, AI customer-support agents, an AI hiring automation platform at Asynq, GyanSathi (a university learning platform for 5,000+ learners with AI-based student evaluation, built solo), SelfAgentic, and carcode.
carcode is an open-source tool that lets you code from your car. You say "Hey Siri, Jarvis" and Claude Code reviews pull requests, reads Slack and ships code changes on your Mac, while Apple CarPlay reads the answers aloud. It is MIT licensed and needs no API key.
Email harshrastogi0603@gmail.com or connect on LinkedIn. He is open to conversations about AI products, agentic systems and engineering leadership.
Experience
From defence infrastructure to generative-AI platforms, building systems that serve real users.
Leading AI product direction and business strategy for Modelia's generative image and video platform — while coding daily. Ideating new tools, automating manual processes, and shipping features that drive measurable business growth.
Engineered the full AI product platform — generative image engine, merchant-facing app, Studio creative tool, and internal systems serving hundreds of merchants generating thousands of AI images daily.
Built core AI-powered hiring automation platform from ground up — intelligent candidate shortlisting, resume analysis at scale, and an autonomous Agentic AI interviewer.
Contributed to Harvester HCI (Hyper-Converged Infrastructure) solution at India's premier defence electronics PSU.
Blog
Writing about production AI systems, system design, and lessons learned from building platforms at scale.
SpaceXAI released an experimental official TypeScript SDK, @xai-official/sdk: Grok text, voice, image and video in one zero-dependency package, plus server-side X search, web search, code execution and remote MCP. Here's working code for each piece and the production gotchas hidden in the README.
I built carcode, an open-source way to code from your car. Say "Hey Siri, Jarvis" and Claude Code reviews PRs, reads Slack and ships code changes on your Mac while CarPlay reads the answer aloud. Here's how it works, what broke on real drives, and how to set it up in five minutes.
On June 30, 2026 the US lifted the export controls it placed on Claude Fable 5 and Mythos 5, and Fable 5 came back online globally on July 1. Here's the full timeline, what changed under the hood — a 99%+ bypass classifier, refusals, and fallback billing — and what developers need to do to ship on claude-fable-5.
SKILL.md files for Claude Code and other coding agents — LLM cost audits, agentic error recovery, AI feature rollouts, and more. Copy, download, or install with one command.
Curated AI developments that actually matter — model releases, agents, MCP, and AI product moves — with takes from the trenches of shipping generative AI.
Get notified when I publish new articles about AI systems, system design, and production engineering.
Skills
I work with modern technologies to build production-grade, scalable systems. From frontend interfaces to cloud infrastructure and AI pipelines.
95%
Escalation Reduction
Platform stabilisation
5K+
Active Learners
University-scale LMS
35
Day Platform Rebuild
Full re-architecture
1000s
AI Images Daily
Production scale
Worked with
I'm always open to interesting conversations about AI systems, engineering challenges, and technology.
or email me at harshrastogi0603@gmail.com