Skip to content
Menu
Interning at RecurlyBroomfield, CO

UtkarshBali

I build agent infrastructure, developer tools, and products people actually use.

Right now I’m a software engineering intern at Recurly. Before that I built agent infrastructure at QualGent (YC X25), and I co-founded Checkpoint, where I’m CTO.

View workGet in touchRésumé
Utkarsh Bali sitting on a hilltop above the coast, smiling into the light as the sun drops over the ocean behind him
4C processPoint to inspect
Education
Purdue University, CS + AI
Graduating
Dec 2026
Based in
West Lafayette, IN
  • 3K+users across 22+ countries
  • <1%task failure rate at scale
  • 45+tools and sub-agents unified
  • 0.844AUROC on hypothesis verification
  • Top 10%of YC S26 applicants
  • ~3xfaster shipping at Recurly
  • 300+students taught
  • ~25%compute spend cut at Microsoft
Now

Building the agent platform that ships Recurly's code.

I built Recurly's core PDLC automation platform: specialized Claude agents (planner, decomposer, a Sonnet/Opus coder pair, and a PR responder) deployed as isolated Kubernetes Jobs through the Agent SDK. It drives a PRD all the way to merged pull requests across five human checkpoints, and it has roughly tripled how fast that path moves.

  • PDLC automation platform: PRD to merged PRs via specialized Claude agents running as isolated Kubernetes Jobs, with five human checkpoints. About 3x faster shipping.
  • AI SDR Slack bot in async Python with 3-tier Claude routing, querying BigQuery, 6sense, ZoomInfo, and SFDC. Returns source-cited briefs and outreach drafts in about 25 seconds.
  • Prospect Report Card pipeline (n8n to FastAPI on Cloud Run to Claude Opus) that turns an inbound form-fill into a personalized PDF analysis and emails it back.
Organisation
Recurly
Role
Software Engineer Intern
Location
Broomfield, CO
Period
May 2026 to Aug 2026
Last updated
Selected work

Agent infrastructure, developer tools, and a few things I shipped end to end.

Each one has a write-up: what the problem actually was, what I built, and what it did or didn’t prove.
Agent reliability2026

Top 10%

of YC Summer 2026 applicants

Agent reliability · 2026

Checkpoint

CI/CD for AI agents. Adversarial test suites that run before an agent ever meets a user.

Read the case study

Interpretability research2026

0.844

AUROC against a synthetic oracle

Interpretability research · 2026

CLIP-H

Clinical hypothesis verification on MIMIC-IV using sparse autoencoders and an LLM ensemble.

Read the case study

Agent infrastructure2025

<1%

task failure rate at scale

Agent infrastructure · 2025

App Crawler

A GKE-distributed crawler that indexes Android apps into an org-wide knowledge base for downstream agents.

Read the case study

All work (7 projects)
Writing

Things worth thinking about twice.

Contact

Building something in this world? Let’s talk.

I’m always up for a conversation about agents, developer tools, or a product you think should exist. The fastest way to reach me is email.

baliutkarsh2@gmail.com