Alchemy

Competitive Intelligence & Product Roadmap

Natural-language image analysis for life-sciences labs.

Company Overview

Alchemy is a local-first desktop app that turns natural-language requests into reproducible life-sciences image-analysis workflows. It serves biotech, pharma, and academic researchers who analyze microscopy images without scripting.

Latest Intel

Zeitgeist tracks private signals to determine where the company is heading strategically.

New Intel: Alchemy is packaging bioimage analysis as a local, agent-built desktop workflow. If scientists trust the verification loop, it can pull budget from ImageJ-adjacent scripting and CellProfiler-style pipeline work.

What They're Building

The company's public product roadmap & what they're committed to building.

Natural-Language Workflow Builder

Researchers describe the image analysis they need, and Alchemy assembles a custom workflow rather than asking users to script or tune every step.

Model Selection Agent

The product claims to select suitable computer-vision models for the task, which moves setup work from the scientist to the software.

Step Verification And Export

Users can verify each analysis step and export reproducible, publication-ready numerical results.

Local-First Desktop Processing

Image processing runs on the user machine, a product choice that fits unpublished research images and sensitive lab data.

Train And Share Vision Models

YC describes Alchemy as helping researchers train and share computer-vision models, pointing toward reusable lab-specific analysis assets.

Founder and Key Execs

Phillip Baek

Founder and CEO (biomedical engineer from UT Austin, previously at Amazon, with research experience at Stanford and UT Austin)

Liam McBride

Founder and CTO (UIUC computer engineer, prior work at Atsign, Stanford University, CivitasAI, and Bobo)

Founder Force Multiplier

Baek brings biomedical research context from Stanford and UT Austin, while McBride brings computer engineering and startup CTO experience. The pairing fits a product that must speak lab language while packaging computer vision for non-coders.

Funding History

2025 | Founded
2026 | Joined Y Combinator Spring 2026, with YC standard terms publicly described as $500K across two SAFEs

Competitors

ImageJ/Fiji:

Open-source microscopy analysis with deep scientific adoption, but less centered on natural-language workflow assembly.

CellProfiler:

Open-source quantitative image analysis for biology, with more manual pipeline construction than Alchemy claims.

QuPath:

Open-source bioimage analysis used heavily in pathology and microscopy, with a different center of gravity than agent-built lab workflows.

ilastik:

Interactive machine-learning segmentation and classification software, closer to model-assisted analysis than end-to-end agent orchestration.

napari:

Python-based image viewer and analysis ecosystem for scientific users, with stronger developer orientation than Alchemy’s no-code surface.

Alchemy

's Moat:

The moat is not proved yet; the path is workflow switching costs from validated lab-specific pipelines and reusable image-analysis models.

How They're Leveraging AI

Natural-language bioimage workflow assembly

Alchemy uses an agentic product layer to turn a researcher’s plain-language request into a custom image-analysis workflow. The user can inspect the steps before exporting numerical results.

Computer vision for microscopy quantification

Alchemy applies computer vision to microscopy and life-sciences images so researchers can count cells, measure motion, and convert images into usable numeric outputs.

Local-first model execution for sensitive research images

Alchemy appears to run image processing and analysis on the user’s local machine, which keeps unpublished lab images closer to the researcher. That design makes AI analysis easier to adopt in labs that do not want raw research data leaving the device.

AI Use Overview:

Alchemy applies an agent layer to translate plain-language scientific intent into computer-vision pipelines, with local execution and human verification around the output.

Bio, Drug Discovery, and Medical Imaging

10xScience

Turns weeks of manual protein analysis into minutes for drug development teams.

CellType

Simulates human biology with AI to predict drug effects and run virtual trials.

Google CEO Sundar Pichai highlighted their C2S-Scale 27B model for generating a novel cancer hypothesis that was later validated experimentally. CellType simulates human biology to predict drug effects using foundation models presented at ICML 2024, with all deals coming inbound from top-10 pharma. Yale professor turned down Google to build it.

Ditto Biosciences

Mines parasite genomes to discover protein therapeutics for autoimmune diseases.

Parasites evolved to suppress human immune responses over millions of years. Ditto mines their genomes for immunomodulatory proteins, and over 98% of those proteins remain uncharacterized. Three PhDs from Harvard, Berkeley, and UCSF with 40+ years of combined expertise in host-parasite biology.

Mango Medical

Automates surgical planning for orthopedic procedures from CT scans using agentic AI.

Surgical planning for shoulder replacements takes hours of manual CT scan interpretation. Mango automates it in seconds via API, already past $500K ARR before demo day, with an 8-figure LOI from a leading orthopedic company. Only company in YC W26 pursuing FDA 510(k) clearance.

Back To All Companies >