"The key takeaway from this is participants feel "empowered" to step outside of their domains and comfort zones to truly learn, build & deliver in less than 24 hours." Luis, Organizer @ Bay Area Lab Automators
With the speed of updates from AI tools, four months can feel like a year. That kind of speed leads to people feeling overwhelmed or left behind without knowing where to start.
Between our last AI + Lab Auto Hack in June and this one, the tools got an upgrade: Potato AI released their Tater agent, new updates to hardware APIs enabling fast starting points, and more MCPs have been built for better integration.
This evolution makes the experience of these weekend hacks more important than ever because they're not just about building projects. They're about pushing the edge of what's possible when small teams get empowered to build + ship in 24hrs with the latest in AI and hardware.
This also may be the first time ever that a group of strangers went from no idea to full experiments with initial cell data in only 24hrs... which is a big deal for advancing science and lab automation as cells don't move at the speed of electrons (ie. software updates).
Let's walk through how we did it and what got built.
The Setup: New AI tools, hardware, and cells
An inspiring convergence of the Bay Area's best minds in AI, science, and hardware.... It was an amazing opportunity to explore how to tackle cell culture in lab automation. Mandana, Researcher @ Plasmidsaurus
AI science tools + hardware (with a real workcell!) + right cell line = path to real experiments
Unlike typical hackathons where teams simulate or prototype only in software, every team had access to Monomer Bio's full robotic workcell and cells as well as more open hardware. This meant teams could design, test, and iterate on actual cell culture experiments throughout the weekend.
Potato AI released their Tater agent with tools to go from idea to scripted protocol autonomously + with human in the loop
AI science partners provided the intelligence layer:
Potato AI for experimental design, literature analysis, and data-driven optimization
Briefly Bio for protocol capture and reproducibility
Cypher AI for creating custom lab software tools without coding
Hardware on Monomer Bio's workcell
Monomer opened their workcell through an MCP interface, letting teams programmatically control incubation, media changes, passaging, imaging, and more. This meant AI agents could directly interact with hardware—no manual intervention required.
Opentrons Labworks Inc. OT-2 for liquid handling
Liconic incubators for maintaining optimal growth conditions
KX-2 robotic arm for plate movement and coordination
Byonoy Absorbance 96 Automate
Tecan Infinite® M Plex, multimode microplate reader
Hardware for open track
Opentrons Labworks Inc. Flex for liquid handling
Cephla's Squid robotic microscope for automated imaging
KX-2 robotic arm for plate movement and coordination
Agilent Automated Microplate Centrifuge
Vibrio natriegens as the fastest replicating cell
Image Source: Jennifer Tsang
Everyone worked with Vibrio natriegens because of how fast it grows: 10min in the right environment with heat shaker and 30min normally. Initially discovered in salty marshes, this bacteria was the selected star of the weekend for speed to growth. Novel Bio provided their NBX Cyclone strain for teams to build.
The agenda and track set up
Once teams were formed after kickoff, they had 24hrs to pick their own track and hardware stack.
Phase 1: Research Biology of Mystery Cell Culture
Learn the biology of this mystery cell culture announced the day of using Potato to answer the following questions:
What media components to use for DOE on media and what dilution ranges to try?
What is a good passaging strategy to start with?
How to pick colonies to select desired cell lines?
What pipetting techniques might affect the growth?
Phase 2: 4 parallel tracks
Optimize: media formulation using Briefly to program and perform serial dilutions on liquid handlers. Materials provided:
Scale: Your mission should you choose to accept is to find the optimal passaging strategy using Monomer. Decide when to passage and to how many wells based on plate reader data. Materials provided:
Pick: Your challenge is to pick colonies. You will have to write software that determines locations of the desired colonies from Cephla’s microscopy images. These locations will need to be sent to the Opentrons liquid handler for colony picking. Cypher will help integrate instruments and manage data and runs. Plates have to be moved manually between the instruments. Materials provided:
Wildcard: Choose your own adventure with Flex, Squid, KX-2 Arm, and Agilent Automated Centrifuge.
Shout-out to Jimmy Sastra, PhD, Carter Allen , and the Monomer Bio for setting these up.
Here are the 8 projects built + shipped in 24hrs
"All working toward one bold goal: building an autonomous scientist that can review literature, develop a comprehensive protocol (thanks to Potato), run the entire experiment in the automated TC room at Monomer Bio!, and analyze the data for the next loop. This is mind-blowing!" Max, Head of R+D at Voyant Bio
1. Sanity Check AI
Problem: Scientists spend too much time wrestling with instrument integration, data management, and analysis instead of doing science.
Solution: A collection of powerful tools that handle lab instrument integration through MCPs, automated data management in ELNs, and AI-assisted analysis—letting scientists focus on what matters.
Team: Aldair E. Gongora , Chikara Oe , Eslam Elshahat, Ph.D. , Etai Sapoznik, Ph.D.
2. Scale Me Maybe
Problem: Onboarding unknown cell strains requires extensive manual trial-and-error to find optimal growth conditions.
Solution: An autonomous system that learns and adapts growth conditions through automated execution on the workcell with AI-driven decision making.
Team: Keltoum Boukra , Srusti Sain , Hanah Rahman , Julie Penzotti
3. VibrioBio
Problem: Vibrio natriegens can double every 10 minutes under optimal conditions, but finding that optimum is difficult with standard media.
Solution: Optimized media formulation for maximizing log-phase growth with minimal metabolic byproducts, all discovered through automated experimentation.
Team: Paula Fogel , Jaee Shah , Michael Lapitan , Rob Learsch
4. ML + LLM Condition Optimizer (Team 7)
Problem: When you have limited data from initial experiments, it's hard to know what conditions to try next.
Solution: Combining machine learning with LLMs to intelligently iterate new experimental conditions based on sparse previous data outputs.
Team: David, Antoine, Greg, Shivam
5. Bomb Digger
Problem: Colony screening is tedious, error-prone, and slow—creating bottlenecks in strain development workflows.
Solution: End-to-end automation of colony screening, removal, and verification through image analysis. Achieves 95%+ accuracy, 90% time savings, and 10x throughput compared to manual methods.
Team: Dale Herzog , Ray, Michael, Yuhan, Sesan
6. Midnight Crew
Problem: Cell culture requires constant monitoring and intervention, limiting experimental throughput and creating reproducibility issues.
Solution: Fully automated closed-loop AI cell growth system that monitors, decides, and acts without human intervention.
Team: Ryan George, Bo-Christopher Redfearn, Armin Foroughi
7. Gork
Problem: Vibrio optimization requires both better media formulations and better ways to query research across multiple sources.
Solution: Improved media conditions for Vibrio growth plus an LLM app that synthesizes information from multiple AI models through APIs.
Team: Jeff, Farhan, Tuzun, Haley
8. Bacter-AI
Problem: Bacterial growth optimization is typically done through slow, manual DOE experiments that don't adapt based on results.
Solution: Closed-loop agentic platform that designs experiments, executes them on the workcell, analyzes results, and autonomously plans the next iteration.
Team: Simon Coelho, Akash Ramachandran, Kanupriya Daga, Mandana Askarizadeh, Max Fereydouni, Vivek Ramaswamy
Thank you to everyone who helped make this event possible!
Hosting partner:
Monomer Bio for opening their office and workcell to all teams, helping design the tracks, support all of the hardware, and for creating the MCP interface that made this level of automation accessible. Jimmy Sastra, PhD has championed these hacks since the first one last year and Carter Allen and the team ensured this event was possible with so 40+ builders over a weekend.
AI science partners:
Potato AI for enabling everyone to get enterprise access for the weekend. Thank you Nick Edwards, PhD and Merrit Savener for supporting hands on experience like this as it enabled teams to move quickly from idea to generated protocol.
Briefly Bio for protocol capture and reproducibility. Thanks Staffan Piledahl for helping teams through the weekend
Cypher AI for supporting the colony picking track with their platform. Thank you Michael T. and Yaoyu Yang .
Hardware partners:
Opentrons Labworks Inc. for the Flex liquid handler and team support. Thank you Shanin Barican for enabling community builds like this one.
Community partners:
Lila Sciences for supporting the AI science community
Worldwide Studios for organizing and co-hosting
@Bay Area Lab Automators for co-hosting and Luis Villa for helping teams through the weekend
SF Hardware Meetup for bringing the broader hardware community
Other supporters
Mohammad Ahsan Fuzail from Lila for mentoring teams, helping with Flex support, and lending a hand the whole weekend.
Hongquan Li and Yan for the Squid robotic microscope from Cephla
Jesse Kriege from West Portal for Byonoy Absorbance 96 Plate Reader
Eric from Novel Bio for the NBX Cyclone cell
Maggie from Tecan for plate reader
We're planning the next one soon. If you want to help support accelerating AI science and the next generation of builders in lab automation, reach out.