Felix Brann

Instacart co-founder Apoorva Mehta is back with Abundance, a hedge fund built on a bold premise: Capital allocation drives the economy. It decides which drugs get developed, which technologies get built, and which ideas survive long enough to matter. Yet today, it still runs largely on human judgment. Abundance flips that model. AI agents generating ideas, underwriting trades, and allocating capital end-to-end. Not AI as a tool, but AI as the allocator. At Eclipse, we believe the biggest opportunities come from rebuilding core systems of the economy. This is what that looks like for capital markets. Apoorva, Laksh Aithani, and Justin Olshavsky have assembled a world class team. Consider joining them: https://abundanceco.com/ More on Abundance's mission from Bloomberg's Hema Parmar: https://bit.ly/4cMxkVI Congratulations Apoorva, Laksh, Justin, and Abundance team!

A little over a year ago we were struggling to answer a few hundred calls a month. To make things worse... we were saying the same thing as eveyone else... "Cut costs. Answer your phones with AI" But then a cohort of customers that asked us the same question... "How do I make our AI better than humans?" We didn't know... so - We trained and fine tuned models that talked more like people - We built A/B testing - We built the most advanced voice AI analytics on the market - We partnered with Groq to get the lowest latency on the market - and 100 other things... Then. Our customers scaled. Slowly at first... then faster by the day. We went from answering a few hundred calls a month, to millions. But not just that... we started beating the best human call center reps. Our engine was working... Improving day over day to deliver better outcomes than the next. One of our customers brought 10M in net new revenue in just three months with Phonely. I didn't think it was possible. And our team went from 5 to 30... in a year. You see... Your website has Google Analytics. Your ads have attribution. Your emails have open rates. Your phone calls? Nothing. Until now. Phonely is the engine that optimizes many businesse's biggest revenue source... the phones. Today I'm incredibly excited to be announcing a $16M Series A lead by Base10 Partners with participation from Y Combinator and customers Etech Global Services, TSA Group - Australia EngageCX and more. And to help spend some of our investors money... If you use a call center or another AI vendor comment "Optimize my calls" and we'll answer your phones free for 30 days. Unlimited calls. White glove onboarding.

Feeling honored to be recognized by Business Insider alongside such inspiring peers. Incredible founders trust me as a partner, and I’m grateful to work alongside them during such an exciting time. I’m also deeply thankful to the team at Alumni Ventures for their mentorship, support, and collaboration. It’s a privilege to do this work, and I’m excited for what’s ahead. Onwards!

🌱 We're hiring Moxxie Ventures! We're looking for an ambitious, early-career investor who loves meeting founders early, building real relationships and spotting lightning before it strikes. Main requirements: high energy, high trust, lots of hustle and obsessed with early-stage technology. We know this is a competitive space and we have a high bar. We’ll pay a $10K referral bonus for the person we hire. 🙏 https://lnkd.in/gB2EXx7a

Back from Council of Medical Specialty Societies Annual Conference 2025. One theme dominated: healthcare desperately needs accurate clinical data, but we're drowning clinicians in documentation burden. Data is useless unless it is accurate, yet the two common "fixes" both fail in the real world. Adding more structured fields increases clinician documentation burden and operational cost. Relying on manual abstraction is expensive, slow, and hard to scale. And oversimplifying measures so they fit structured data undermines the whole purpose of clinical registries, stripping out nuance needed for research, risk adjustment, and subgroup analyses. Pathology makes this tension especially stark because the clinically relevant concepts live in messy narrative reports. Defining an initial cohort from billing codes creates huge false positives. Take CPT 88305, a "single gross and microscopic tissue exam" code spanning everything from colon polyps to skin punches to prostate cores. A patient biopsied years ago for prostate tissue can later be swept into a skin melanoma measure unless a skilled reviewer manually rules them out. Until now, that manual exclusion step was the only reliable fix for orgs like the College of American Pathologists (CAP) Another layer of noise comes from uncertain and highly variable language in pathology reports. Pathologists hedge when morphology is borderline or tissue is limited ("suspicious for," "cannot exclude," "favor"), and those nuances matter for measure eligibility. The same clinical concept is expressed dozens of ways across institutions and individuals. Without robust normalization, keyword and rule-based extraction either miss true cases or amplify false positives. Some measures become so brittle that they are effectively intractable; CAP previously had to abandon automatic collection of an intestinal metaplasia measure because existing approaches could not handle the narrative complexity. At CMSS we presented an AI system with explainable, auditable reasoning that preserves clinical nuance while automating abstraction end to end. Across four CAP measures (about 2,000 reports), the system matched human performance, with low-variance outputs and fully reviewable reasoning for each decision - enabling complex measures that were previously intractable. We are grateful to the CAP, and to the abstractors and managers at Kiser Healthcare Solutions, LLC who built the gold labels. Special thanks to Dr Colleen S., who co-presented this work with Felix Brann and myself and did a ton of work getting the dataset and measures ready. Also special thanks to Dr Helen Burstin for organising such an amazing conference! We're scaling this work. If you're a physician AI researcher who can code and wants to join Pharos as a founding clinical AI research team member in San Francisco—building auditable systems that automate quality reporting and help hospitals identify and prevent avoidable harm at scale—please reach out.