Loading profile
Loading profile
· Projects
10 projects — each one is a separate landing page, links out below.
· AI surfaces
Each card is the actual opening of a full AI-written page about Sarthak. Click in.
I am building Karte as a personal presence product: part Linktree replacement, part personal website, part personal chatbot, with shareable generated profile modes layered on top.
Contact
Use a verified DM if you want a real reply. For public context, start with the projects and links on this page.
Blog
Post 01
May 2026
Notes on replacing link-in-bio pages with personal sites, chat, and shareable profile modes.
Visit website->
Post 02
Apr 2026
How memory, links, projects, and public context can compound into a better personal website.
Visit website->
Post 03
Apr 2026
A short argument for profiles that answer questions instead of only listing links.
Visit website->
Experience
Software Engineer@ VaultWealth (Peak XV)
Feb 2025 — Present
Backend services, durable workflows for financial planning. Cut unexpected failure rate ~90%. Building AI-assisted product surfaces.
Backend + Data Engineer@ Front.Page (YC S21)
Jan 2022 — Jan 2025
Real-time market pipeline scaled 15k → 200k DAU; tail latency 600ms → 60ms. BERT/Milvus vector feed (+40% engagement). RAG support agents cut human load 90%.
I build the rails AI runs on — and the products that ride them. Backend + AI infra at VaultWealth (Peak XV). Previously Front.Page (YC S21). Building Karte, CodeVetter, SaaS Maker, free-ai, high-signal, tinygpt on the side.
Software Engineer | VaultWealth (Peak XV) | Feb 2025 — Present | Backend services, durable workflows for financial planning. Cut unexpected failure rate ~90%. Building AI-assisted product surfaces. Backend + Data Engineer | Front.Page (YC S21) | Jan 2022 — Jan 2025 | Real-time market pipeline scaled 15k → 200k DAU; tail latency 600ms → 60ms. BERT/Milvus vector feed (+40% engagement). RAG support agents cut human load 90%.
Spotlight
Interested in AI infrastructure, developer tools, or personal presence products? Let’s talk.
Link unavailable
◆ Timeline
May 2026·Shipped
PyTorch model compiled to WebAssembly + WebGPU. Trains and runs entirely client-side.
May 2026·Launched
Calling card profiles with four AI surfaces (chat, encyclopedia, newspaper, roast) on top of a Cloudflare Workers / D1 / free-ai stack. The page you are reading.
April 2026·Shipped
Evidence-backed, versioned signal log for AI infra / semiconductors. Research artifact first; product layer later. NLP pipeline scoring SEC filings, news, and Reddit.
February 2026·Shipped
Feedback, waitlist, testimonials, changelog, analytics, forms, roadmap, AI gateway, chatbot. The infrastructure layer behind several solo products.
February 2026·Shipped
OpenAI-compatible API router across free providers. Powers Karte's free tier and several other side projects.
November 2025·Shipped
Static-analysis + LLM reasoning over a PR. Desktop client because the SaaS path was too noisy.
October 2025·Shipped
Built reliability primitives around Temporal — retries, idempotency, downstream-failure handling.
February 2025·Joined
Peak XV-backed. Building financial planning systems + reliability infra for the UAE market.
September 2024·Shipped
Vector store + retrieval + grounded generation, integrated into the existing support flow.
January 2024·Shipped
BERT embeddings + Milvus vector search replaced the heuristic ranking. A/B confirmed.
June 2023·Shipped
Tail latency 600ms → 60ms. Rewrote the hot path on Kafka + Go; ClickHouse for analytics.
January 2022·Joined
Fintech social platform. Spent 3 years on the backend, real-time pipelines, vector feeds, and RAG.