Bibek Shrestha

AI Architect at ThoughtSpot

I design enterprise AI deployments at ThoughtSpot and build the agents and tools that make customer delivery faster.

Email me Résumé

Every node is something I built. Click one.

of customer-facing solutions engineering and architecture
7 years
MCP servers scored by MCPpedia
50,000+
to build a ThoughtSpot custom chart with Chart Skills, down from days if not weeks
Minutes
  • Claude Code
  • Model Context Protocol
  • Cloudflare Workers
  • Next.js 16
  • Supabase Edge Functions
  • pgvector
  • Gemini
  • Playwright
  • Neon Postgres
  • FastAPI
  • Zod
  • Snowflake
  • BigQuery

Selected work

MCPpedia home page with server search and scores

Open catalog and MCP server

MCPpedia

Before you install an MCP server, you want to know if it is secure and still maintained.

I built an open catalog that indexes MCP servers and scores each one on security, maintenance, efficiency, documentation and compatibility, with real CVE scanning and tool-poisoning detection. It also ships as an MCP server on npm (7 tools, stdio and Streamable HTTP), so you can search and compare servers from inside an AI client.

50,000+ servers scored. Listed on Smithery.

mcppedia.org ↗MCP server repo ↗

Photobari home page with a full-resolution sunset photograph

Photo platform, live in production

Photobari

I wanted to share photos at full resolution, with no account needed to view them.

I designed, built and launched it solo in about 226 commits, a photo portfolio and private sharing platform where images are never used for AI training. Next.js 16 on Cloudflare Workers, Neon Postgres with Drizzle, Better-Auth, and presigned direct uploads to R2 with BlurHash placeholders and EXIF extraction.

Live in production.

photobari.com ↗My photography ↗

Pasa home page with the early-access sign-up

AI trip planner, in private beta

Pasa

Plan a trip by talking about it, then get a day-by-day itinerary on a map.

Claude gathers trip details in chat, then Gemini with Maps grounding generates an itinerary that is geocoded and route-optimized. Past trips feed back in through RAG: gemini-embedding-001 vectors in pgvector, retrieved by cosine similarity into both prompts. Every AI call goes through Supabase Edge Functions so keys stay server-side, and every LLM JSON response is validated with Zod.

I also built an eval framework for it: 55 scenarios across 30 categories (161 graded checks), graders calibrated against a hand-graded gold set, adversarial-user tests, and a baseline-diff gate that blocks prompt regressions.

5,982 tests passing, 0 type errors. About 1,220 commits and 268 merged PRs. In private beta.

mypasa.ai ↗

More I have built

Claude Code skill

ThoughtSpot Custom Chart Skills

Custom Chart Skills site showing a rendered Sankey chart

A Claude Code skill that builds ThoughtSpot custom charts end to end. It renders each chart in a real browser through Playwright and critiques its own output until the render matches, so customers can build their own charts.

Building a custom chart went from days, if not weeks, to minutes.

Website ↗Repo ↗

Claude Code plugin, open source

agent-org

agent-org site with the install commands

An open-source Claude Code plugin that generates a CEO-led agent org for any repo: tailored agents, an orchestrator, memory, guard hooks, evals, and an autonomy level the org earns on scorecard evidence. Only builder agents can edit code, and every miss a human catches becomes a test the agents must pass.

Adopted in Photobari: 29 merged agent PRs, 0 reverts.

Website ↗Repo ↗

Web app for ThoughtSpot admins

ThoughtSpot Admin Toolkit

Admin Toolkit site with a dashboard preview

A web control plane for ThoughtSpot administrators, with a Python/FastAPI backend and a Next.js frontend. It shows the complete data lineage of any object, from warehouse tables to Liveboards, and puts a web UI on a CS tool that used to be terminal only, so non-technical users can run it. Bulk archive, delete, share and ownership-transfer jobs run with dry-run previews and a TML backup first, and it installs with one line.

Creator and lead developer, 68 of 92 commits, shaped by feedback from field users.

Website ↗Repo ↗

Embed Playground app with embed options and SDK panels

SDK sandbox with a token service

ThoughtSpot Embed Playground

A zero-build sandbox for the ThoughtSpot Visual Embed SDK, with a hardened Express service that mints trusted-auth tokens server-side. A multi-agent engineering org maintains it, and a PR merges only when CI is green and review finds no confirmed bug.

86 commits. Security-critical changes are held for human approval.

Live demo ↗Repo ↗

JavaScript chart library on npm

mayaCharts

mayaCharts site with an animated dot chart

An open-source JavaScript chart library: 28 chart types from one JSON spec, about 87 KB gzipped, zero runtime dependencies, accessible and server-rendered by default, and able to handle up to 1M rows.

Published on npm as mayacharts.

Website ↗Repo ↗npm ↗

At work

ThoughtSpot

AI Architect, Customer Outcomes 2026 to now

Solutions Architect 2025 to 2026

Across many different customer use cases, I look for the slow step in how we deliver and build the tool that removes it.

  • Embed Playground. I built it to cut validation of customer embedding requirements from days to hours, and to turn call explanations into live, working demos.
  • Admin Toolkit. I built it (68 of 92 commits) with lineage graphs that show the complete picture of data lineage, down to the column, and a web UI that lets non-technical admins run a terminal-only CS tool.
  • Custom Chart Skills. My open-source Claude Code skill cut building a custom chart from days, if not weeks, to minutes, so customers can build their own.
  • AI architecture. I own enterprise AI architecture for customer deployments in retail, financial services, healthcare and SaaS: embedding, semantic modeling, agent APIs and security (row-level security, ABAC, multi-tenant isolation).
  • AI-assisted delivery. I build AI agents, agent tooling and Claude skills that replace slow, repetitive steps in the field team's delivery, and I set the team's pattern for adopting AI.
  • Codex Award. I won ThoughtSpot's peer-nominated Company Codex Award for the value "We measure time in hours and days, not weeks and months" (SKO, Aug 2026).

Toucan

Lead Solutions Engineer 2023 to 2025

Solutions Engineer 2021 to 2023

Technical Customer Care 2021 to 2022

  • $1.03M in new revenue. I led proofs of concept for 100+ prospects, designing a tailored BI solution for each.
  • Technical discovery. I ran deep dives with Account Executives on prospect calls, mapping each prospect's requirements and what Toucan could and could not solve.
  • 500+ dashboards across health, finance, retail and eCommerce.

How I build

I build with Claude Code, and I run it with the discipline I would expect from any engineering team: a spec first, automated tests, CI gates, adversarial review, and human approval on anything security-critical.

  1. Spec first
  2. Automated tests
  3. CI gates
  4. Adversarial review
  5. Human approval

My background is seven years of customer-facing solutions engineering and architecture, and I have not held a software engineer title. So I let the projects above carry the proof, and most of them are public or live so you can check them yourself.

  • Pasa ships behind 5,982 tests and a baseline-diff gate that blocks prompt regressions.
  • In agent-org, reviewers judge a frozen tree and every miss a human catches becomes a new test.
  • The Embed Playground merges a change only when CI is green and review finds no confirmed bug.

↑↓ move↵ openesc close