Case Study

Developing AI-Driven Biomass Monitoring for Remilk's Fermentors

About

Remilk, a global innovator in animal-free dairy production, leverages precision fermentation to produce dairy-identical milk proteins. Through its patented technology, Remilk significantly reduces resource consumption compared to traditional dairy farming and completely eliminates dependence on dairy cows in industrial-scale production. Remilk’s dairy products deliver the familiar taste and texture of traditional dairy without lactose, cholesterol, hormones, or antibiotics.

Challenge

It was critical for Remilk to accurately measure biomass within their fermentors to ensure successful fermentation and cell culture process. They were using traditional biomass assessment methods that were resource-intensive, sporadic, and often yielded non-representative samples.

Time-Consuming Sampling

Sampling for cell density can take around 30 minutes per event, with additional analysis extending this into multiple hours weekly.

Latency in receiving results

Due to the lag between sample collection and result delivery, often several hours, made it difficult to catch anomalies early, increasing the likelihood of undetected issues that could impact batch quality and yield.

Error-Prone Process

Manual methods like manual cell counting, dry weight determination, or offline OD measurements introduce variability due to factors like sample handling, dilution accuracy, timing, and subjective interpretation.

Remilk required a continuous, accurate biomass measurement solution integrated seamlessly into their fermentation processes.

Solution

BioRaptor developed an AI-powered virtual biomass sensor (V-sensor) capable of continuously predicting cell density in real-time. The sensor utilizes existing fermentor sensor data and historical biomass measurements, trained through advanced machine learning algorithms. Key functionalities include:

Unified Data Collection

Remilk’s scientists use BioRaptor's platform to collect online, offline, and historical experimental data. Ingesting and collecting the data is easy thanks to a multitude of data connectors. Data is collated from both online instruments and ‘offline’ sources such as excel files, CSVs, and other places (e.g. Excel, Powerpoint, Electronic lab notebooks).

Insight Generation

Everyone on the team, including scientists, engineers, and managers, can easily access and query the collated data to discover patterns within subsets of the body of experiments. Scientists and engineers can superimpose multiple experimental parameters and sensor readings to find patterns visually.

Predictive Modeling

The data modeling module enables a code-free experience, where scientists and engineers can drill down into complex time series data, create models or train pre-built models to predict outputs upon changing specific parameters (thus saving precious resources such as time, budget and lab equipment).

Automated Analysis

The V-Sensor automates key analytical tasks, such as biomass estimation, metabolic activity tracking, and specific metabolite measurements. Trained on both historical online and offline data, the V-Sensor builds a model that enables real-time estimations during the ongoing process. This solution not only streamlines analysis but also replaces lengthy, costly, and labor-intensive manual measurements.

Real-Time Monitoring

AI-driven predictive analytics enable real-time, remote monitoring, allowing for the immediate detection of process deviations and issues thus preventing costly failures. Additionally, customizable alerts, whether triggered by process deviations from setpoints or by AI-detected anomalies, are sent via email or text message to alert teams and improve decision-making.

BioRaptor automates data collection across bioreactors, harmonizes and contextualizes to align all parameters with BioHarvest’s nomenclature.

fwefewfwe
fewfwefewf
fewfwefweff
fewfwefwfew

BioRaptor automates data collection across bioreactors, harmonizes and contextualizes to align all parameters with BioHarvest’s nomenclature.

fwefewfwe
fewfwefewf
fewfwefweff
fewfwefwfew

BioRaptor automates data collection across bioreactors, harmonizes and contextualizes to align all parameters with BioHarvest’s nomenclature.

fwefewfwe
fewfwefewf
fewfwefweff
fewfwefwfew

Results

Instantaneous biomass measurement - cell density monitoring is reduced from hours to microseconds, providing immediate, continuous insights to promptly detect and address fermentation issues.

Resource optimization - BioRaptor’s AI-driven V-sensors facilitated real-time adjustments and process optimization, leading to enhanced yields and improved product quality while significantly lowering production costs.

Enhanced efficiency - reduced reliance on manual sampling freed scientists to focus on critical aspects of experimentation and innovation.

“The amount of data we collected from experiments is extensive, making it difficult to analyze everything. We are using BioRaptor to identify patterns by combining different combinations and then to make predictions about future experiments. It allows us to gain a far deeper understanding of the phenomena we are studying and identify new paths to investigate, really allowing us to move faster and with greater confidence.“