About

Khushal Sahni

Engineer who owns the problem: products, backend systems, and AI agents

I design distributed systems, ship products end to end, and build agent workflows that hold up in production. I care about the economics of what I build, not just the architecture.

Open to full-time roles, founding teams, and high-stakes buildsGurugram, IndiaIIT (BHU) '22
ResumeEmailLinkedIn

What I bring

Backend systems

Payments, checkout, recommendations, and microservices at Urban Company scale. Event-driven designs, explicit failure modes, and ownership of money paths.

Checkout note →AniReco systems →

Products end to end

Vivaahli: marriage biodata and digital invites. Product, UI, payments, PDF pipeline, and go-to-market. Live consumer product with real families using it.

Vivaahli case study →Ownership note →

AI agents

Multi-agent orchestration and tool-calling workflows in production SaaS. Not ML research: shipping agents that call tools, fail safely, and get evaluated.

Agents note →Zashiki Warashi →

What I take / what I skip

Good fit

  • Full-time roles where I own a surface end to end
  • Founding or early engineering teams shipping a real product
  • Backend or full-stack work with production stakes (payments, agents, consumer)
  • Problems where judgment and shipping beat ticket velocity

Usually a pass

  • Pure ML research or model training without a product
  • Roles that are only CRUD and meetings
  • Agencies that need a pair of hands, not an owner

Track record

SDE 2 · Fello

Feb–May 2026

Building backend infrastructure from the ground up: communication systems, AI integrations, and multi-agent orchestration at an early-stage SaaS.

SDE 1 → SDE 2 · Urban Company

~4 years

Backend and full-stack work on payment systems, checkout, recommendations, and microservices at scale, through IPO readiness.

Re-architecting checkout →

Indian Institute of Technology (BHU) Varanasi

B.Tech · 2022

Selected work

  • Vivaahlilive

    Wedding-tech for Indian families

    Biodatas created: 1.4K+ · Languages: 10 · Family rating: 4.8/5 · Time to finish: ~5 min

  • ourmoneylive

    Where did the reported rupee for my place go?

    Hackathon: Build What Moves India · Shortlisted from: ~5,000 · Data: Mock + live extract · RTI: Drafts, never files

  • Zashiki Warashilive

    A house spirit for the localhost pile

    Platform: macOS-first · Catalog: Machine-local SQLite · Rust tests: 44 · License: MIT

  • AniRecoarchived

    Personalized anime recommendations from your AniList

    Hybrid ranker: 5 signals · Retrieval: 3 vectors · Catalog: 50k+ anime · Outcome: Archived

Next step

Read the resume, skim a case study, or email me directly.

ResumeWork[email protected]