Sarthak Agrawal is an AI infrastructure and product engineer known for building durable systems and consumer-facing AI tools. His work spans backend services, reliability engineering, and AI-native products such as personal presence platforms, code review assistants, and developer tooling. Agrawal’s public projects emphasize practical AI deployment, from LLM gateways to vector-powered workflows, and he describes himself as focused on turning context into useful software surfaces.

Overview

Agrawal’s public profile presents him as a builder of AI infrastructure and products, with a stated preference for shipping consumer-facing AI or developer-facing infrastructure rather than pure research roles.^[S7] His engineering philosophy emphasizes handling edge cases—timeouts, retries, idempotency, and downstream failures—over idealized happy-path code.^[S21] Public descriptions highlight a technical style centered on full-stack systems, including Next.js, Cloudflare Workers/OpenNext, D1/Turso, Clerk-gated dashboards, and AI-assisted workflows.^[S17]

Career

Agrawal has worked in backend and data infrastructure roles at venture-backed technology companies. He joined VaultWealth, a Peak XV-backed financial planning platform, in February 2025, where he builds backend services and reliability infrastructure for financial workflows in the UAE market.^[S32] Previously, he spent three years at Front.Page (YC S21), a fintech social platform, leading backend and data infrastructure efforts.^[S36]

At Front.Page, Agrawal scaled a real-time market data pipeline from 15,000 to 200,000 daily active users while reducing tail latency from 600ms to 60ms.^[S35] He also implemented a personalized feed using BERT embeddings and Milvus vector search that increased engagement by 40%, and deployed RAG-powered support agents that reduced human support load by approximately 90%.^[S34,S33]

Notable projects

Agrawal has released multiple open-source and commercial products, often documented through public repositories and project websites:

• Karte — An AI-native personal presence product combining a link-in-bio page, personal website, personal chatbot, verified or anonymous DMs, and generated profile modes such as Encyclopedia, Newspaper, and Roast Me. The system uses a Profile Memory layer to power visitor chat and structured profile pages.^[S37,S5,S14]
• CodeVetter — A desktop-first review platform for AI-generated code that performs static analysis and LLM reasoning over pull requests to catch vulnerabilities, regressions, and silent drift before code ships.^[S38,S30]
• SaaS Maker / Fleet — A developer tooling system for launching and operating SaaS products, offering features such as waitlists, feedback, testimonials, changelogs, analytics, roadmaps, and an AI gateway.^[S39,S28]
• High Signal — A product workspace for finding and acting on evidence-backed signals, including NLP pipelines that process SEC filings, news, and Reddit for AI infrastructure and semiconductor domains.^[S40,S27]
• Starboard — A repository discovery and search tool with natural-language query support, hybrid lexical/semantic retrieval, and weekly digest reporting for high-signal repositories.^[S41]
• SignificantHobbies — A consumer product for mapping a person’s hobby journey across life phases and sharing a timeline of interests.^[S42]
• RolePatch — An AI-powered resume-tailoring and job-fit tool with ATS scoring, contextual rewriting, and interview preparation features.^[S43]
• TinyGPT — A 0.8M-parameter transformer that trains and runs entirely in the browser via PyTorch compiled to WebAssembly and WebGPU, intended as a teaching artifact demonstrating the minimal scale of a working GPT.^[S45,S8,S25]
• free-ai — An OpenAI-compatible LLM gateway running on Cloudflare Workers that routes to multiple providers such as Cloudflare Workers AI and OpenRouter, used to power Karte’s free tier.^[S46,S9,S29]

Online presence

Agrawal maintains public profiles on several platforms, where he describes himself as open to collaboration and AI infrastructure or product roles. His primary website is sarthakagrawal.dev, which highlights his focus on LLM gateways, real-time pipelines, vector search, and durable workflows.^[S48,S52] His GitHub account (@sarthakagrawal927) hosts over 120 public repositories, including many of the projects listed above, and features badges such as Pair Extraordinaire and Pull Shark.^{[S23,S49,S52]} He is also active on Twitter (@sarthakcodes), where he shares builder-focused updates and expresses openness to collaboration.^[S51,S15] A LinkedIn profile is also listed publicly.^[S50]

Public profile and collaboration

Agrawal’s public profile indicates openness to AI infrastructure and AI product roles, including full-time and fractional engagements, with a preference for teams shipping consumer-facing AI or developer-facing infrastructure.^[S7] He suggests that inquiries begin by reviewing his projects and links, and recommends using the calendar booking link on his profile for a 20–30 minute call, accepting async communication via email when preferred.^[S12] His public X profile explicitly states he is open to collaboration.^[S15]

Technical approach

Agrawal’s technical work emphasizes production-grade systems and practical AI deployment. His stack includes Go, Node.js/TypeScript, Python, Kafka, Temporal, Docker, Kubernetes, and Socket.io on the backend, with RAG pipelines, OpenAI APIs, BERT embeddings, and Milvus vector search for AI components.^[S22] Frontend work often uses React and Next.js, with Cloudflare Workers and D1/Turso for data and deployment.^[S17,S22] His public commit messages reflect a preference for concrete constraints, verification, and iterative improvement in production environments.^[S17]