I help organizational leaders generate ROI from their AI investments, particularly in healthcare and life sciences, by building systems to incorporate domain expertise, leveraging pragmatic evaluations, and embedding LLM-native best practices. My experience:

• 9 years practicing and training as a doctor, 7 years building AI products. My healthcare AI research has 1700+ citations (including the first review paper comparing AI to clinicians). I trained ML algorithms pre-LLMs and then helped write the playbook for LLM-powered applications in healthcare.
• Head of Clinical AI at Sequoia-backed startup Anterior (seed -> series B), founding AI lead at Cera (seed -> series B) and founder of an EF-backed healthcare startup.
• Built healthcare AI systems processing millions of patients with 99%+ accuracy, pioneering novel LLM architectures and human-in-the-loop review processes that scaled from solo development to production teams of 20+.

I love working with frontier healthtech and life science startups (Series A and beyond) to build production-ready AI systems from the ground up, and with healthcare enterprises - including providers, payers, pharma, and medtech companies - to turn AI pilots into deployed products that deliver measurable value.

Building AI products today requires a fundamentally different approach

95% of generative AI pilots deliver zero return. I see three critical challenges holding teams back:

1. The MVP-to-production chasm

Demos are easy - production ROI is hard. You need to solve the last mile problem: AI systems work well in general but fail on your specific workflows without deep domain expertise and operational context baked in. The bar for performance in mission-critical industries is high and not easy to achieve.

I help teams design pragmatic evaluations, tailored to their stage of product development, to enable targeted iteration, achieve safe and effective performance and then maintain it through robust post-deployment monitoring. At Anterior, I built AI systems that achieved 99.24% accuracy and 92% clinician satisfaction across millions of patients. I made this possible by designing human-AI hybrid review systems, enabling a small team of clinicians to review cases from millions of patients.

2. Domain expertise matters more than ever

The vertical AI imperative is here. Vertical AI captured $3.5 billion in 2025, nearly tripling year-over-year, while horizontal AI solutions struggle to access the domain-specific data and regulatory knowledge that create defensibility. The market has spoken: domain-specific AI wins, generic AI doesn’t ship.

The problem: domain experts and engineers speak different languages. Clinical teams can articulate what “good” looks like but can’t translate it into system requirements. Engineering teams can build sophisticated systems but don’t know which clinical nuances actually matter. Traditional software workflows weren’t designed for this.

I bridge this gap. As a doctor who’s built AI systems for 7 years, I understand both worlds. I can help tech leaders architect LLM-native systems, clinical teams prove performance for regulatory and safety requirements, and business leaders demonstrate measurable ROI from AI investments.

I’ve designed workflows, tools and processes that convert domain expertise into production-ready LLM applications - turning institutional knowledge into prompts, evaluation methods, and system architecture that performs.

3. Companies are building with yesterday’s playbook

The “pre-LLM” era approach doesn’t work in today’s LLM world, yet most companies haven’t caught up. Building LLM-powered products requires new skill sets and approaches: context engineering, workflow decomposition and evaluation systems - rather than feature engineering, model architecture design and model accuracy optimization.

I’ve lived through this evolution. I operated in this ‘pre-LLM’ world; I built state-of-the-art deep learning algorithms and published the first study to compare AI and doctors. I then spent 2 years building LLM-powered products for healthcare admin at a Sequoia-backed start-up. I help teams build in an LLM-native way; in how they architect their applications and write code, in how they hire and train their staff and in their operational processes.

Let’s discuss your AI challenges

I’m currently accepting a limited number of new consulting engagements. If you’d like to explore whether we’re a good fit for each other, book a call below:

Book a call