Triangle.AI is developing a platform to revolutionize how patients find and engage with clinical trials.
“Spartan’s founder, Chan, is one of the best engineers I’ve ever worked with—I used to joke that I wanted to figure out how to clone him. With Spartan, it feels like he’s done exactly that. Their hiring bar is incredibly high, and the developers I’ve worked with are all at the same high level, consistently delivering outstanding work with impressive speed.”
- Justin Balthrop Co-founder & CTO of Triangle, Co-founder of Bird @ Triangle, Bird
Overview
Triangle is a Pasadena-based startup transforming how patients and providers discover clinical trials. The company is building an AI-powered search and matching engine that dramatically outperforms clunky government databases like ClinicalTrials.gov.
The product leverages LLMs to parse trial data, simplify eligibility criteria, and connect patients to the right studies in seconds instead of weeks. Triangle’s mission is to remove the friction from trial discovery and accelerate patient enrollment globally.
By mid-2024, Triangle was racing to ship a working MVP to demo for investors, partners, and major medical orgs. That’s where Spartan came in.
Spartan’s Solution
Spartan deployed a squad of 2 AI engineers + 1 fullstack engineer, operating as an embedded product team alongside Triangle’s founding team.
Execution highlights
- Built a zero-shot eligibility matching engine using LLMs → patients can match to trials without pre-labeled data
- Developed an AI-powered search interface + chat interface that drastically outperforms ClinicalTrials.gov
- Implemented an operational dashboard so Triangle could track how well the AI was performing end-to-end
- Designed and deployed an Airflow-style data pipeline powered by LLMs → transformed raw trial data into a structured database for fast querying and analysis
- Created a page generation pipeline that combines scraped clinical trial data, PubMed insights, and LLM-generated summaries into user-friendly trial pages
- Authored technical RFCs to scale the matching algorithm and guide future system improvements
- Supported privacy-first features like anonymous patient profiles and multilingual support
The team shipped iteratively with daily standups, async Slack communication, task tracking in Notion, and continuous deployment through GitHub.
Before vs After Spartan
| Area | Before Spartan | After Spartan |
|---|---|---|
| AI Eligibility Matching | Manual, error-prone, non-existent | 87.3% accuracy → near expert-level matching |
| Search UX | Government-grade → confusing + unusable | AI-powered search + chat UX → intuitive + fast |
| System Observability | No visibility into model or system quality | Full dashboards + feedback loops on model accuracy |
| Trial Discovery | Text-heavy pages, hard to parse | Automated trial pages with LLM summaries + metadata |
| Screening Speed | Manual, slow | 42.6% reduction in trial screening time |
| Data Pipelines | Scraped JSON → brittle | Clean database with structured, enriched trial data |
| Dev Velocity | Blocked, internal interns only | MVP shipped in under 12 weeks with Spartan team |
The Outcome
Triangle moved from concept to live demo in record time.
- 400,000+ clinical trials searchable → global coverage across trial databases
- 87.3% matching accuracy → validated against expert benchmarks
- 42.6% faster screening → reduced patient-to-trial matching time drastically
- Multi-language support + privacy-first features unlocked broader markets
- Demoed the product to key partners like Cancer Commons, leading investors, and billionaire donors in healthcare
The MVP positioned Triangle to kick off fundraising conversations and secure partnerships with leading clinical orgs.
Why Spartan?
Triangle didn’t just need engineers. They needed a build-partner who understood AI, product, and scale, someone who could help them think through the entire system while shipping fast.
Spartan delivered because we:
- Operate like co-founders, not contractors → RFCs, product input, system design
- Move at startup velocity → MVP in 12 weeks, not 6 months
- Bring real AI expertise → not just GPT wrappers, but production-grade AI infra
- Pair engineering craftsmanship with pragmatic delivery → built for now, but scalable for what’s next
- Cost 50–60% less than SF or NY hiring → with better output
This is why Spartan clients keep coming back.
What’s Next
- Scaling the AI matching engine to support rare disease trials and global datasets
- Adding patient outcome tracking to improve model recommendations over time
- Partner integrations with clinical orgs, hospitals, and research foundations
- Moving from MVP to full GA release with expanded team
Triangle’s partnership with Spartan set the foundation for transforming how the world accesses clinical research.
