Real answers about cell therapy manufacturing, GMP, data systems, and careers — from someone working at the intersection of all of them.
Try different words, or
Most medicines are chemicals — you manufacture them in a factory, bottle them up, and ship millions of identical doses. Cell therapy is fundamentally different. The medicine is living cells — usually taken from a patient or a donor, modified or expanded in a lab, then returned as a therapeutic product.
Because you're working with living biological material, every batch behaves slightly differently. Cells respond to temperature, timing, and handling. A conventional drug doesn't care if it sits on a shelf — cells do. That's what makes cell therapy one of the most technically demanding environments in modern medicine, and honestly, one of the most exciting.
CAR-T stands for Chimeric Antigen Receptor T-cell therapy. In plain language: a patient's immune cells (T-cells) are collected, genetically engineered in a lab to recognize and attack cancer cells, then infused back into the patient.
It's a big deal because for some blood cancers — like certain leukemias and lymphomas — it's produced complete remissions in patients who had no other options. The FDA has approved several CAR-T products including Kymriah and Yescarta. The challenge: each dose is made from that specific patient's cells, so manufacturing is incredibly personal, complex, and expensive.
Autologous means the cells come from the same patient receiving treatment. You collect their cells, process them, and give them back. Highly personalized, avoids immune rejection — but expensive, slow, and every batch is unique.
Allogeneic means cells come from a healthy donor and can theoretically treat multiple patients — more like a traditional "off-the-shelf" product. More scalable, but introduces immune compatibility challenges. Most approved CAR-T therapies today are autologous, but allogeneic is a major research frontier right now.
Living cells can't survive indefinitely at room temperature — or even in a regular freezer. Liquid nitrogen (LN₂) maintains temperatures around −196°C, which essentially pauses all biological activity. Cells can be stored for years without significant degradation at this temperature.
In a manufacturing environment, cryogenic tanks hold thousands of samples in precisely labeled racks and locator slots. Tracking which slots are occupied versus available is a real operational challenge — one I've built systems to solve. Every sample needs a full chain of identity: where it came from, where it went, and exactly where it sits right now.
In a cell therapy facility, you're dealing with enormous amounts of operational data — equipment readings, batch records, environmental monitoring, quality events, inventory, scheduling. "Data systems" refers to the software infrastructure that captures, stores, validates, and surfaces all of that.
The key systems: MES (Manufacturing Execution System — tracks what happens on the floor), LIMS (Laboratory Information Management System — manages lab samples and test results), and QMS (Quality Management System — handles deviations and compliance). Getting these to talk to each other cleanly is a significant, ongoing technical challenge.
EBS stands for Oracle E-Business Suite — an enterprise resource planning (ERP) platform used by many large organizations to manage inventory, purchasing, financials, and operations. In a manufacturing setting, EBS tracks what's on hand, where it is, and what's been consumed.
My CryoTrack dashboard was built specifically to work with daily EBS exports. Every day, operators generate an on-hand inventory report from EBS, and my system ingests that file to calculate which cryogenic storage slots are available. EBS is built for broad inventory management — not fine-grained lab operations — so a custom layer on top makes a real difference.
Mostly reality — but with important nuance. AI is genuinely useful for anomaly detection in process data, predicting batch outcomes from early signals, and prioritizing quality review workloads. These aren't futuristic — they're starting to be implemented now.
Where it's still more hype: autonomous decision-making in GMP environments. Regulated manufacturing requires human accountability and documented rationale for every quality decision. AI can surface insights, but a qualified person still has to make the call. My research publication covers exactly this — how AI augments rather than replaces human oversight.
Spreadsheets are great for analysis and one-off reporting. They break down quickly when multiple people update the same data, when you need to track history over time, or when the dataset grows large. In an environment with hundreds of tanks, thousands of samples, and daily updates — a spreadsheet becomes a liability.
A relational database like PostgreSQL lets you query large datasets instantly, enforce data integrity rules, store historical snapshots, and serve data to a web interface in real time. It also enables auditing: you can see exactly what changed and when. In regulated environments, that's not optional — it's required.
GMP — Good Manufacturing Practice — is the regulatory framework governing how medicines are made. In the US, it's enforced by the FDA. The core idea: manufacturing processes must be consistently controlled and documented so that the product is safe, pure, and effective every single time.
In cell therapy specifically, GMP matters even more than in conventional pharma. You're working with patient-derived material, using aseptic techniques, producing a product that can't be terminally sterilized. One contamination event or identity error isn't just a quality problem — it can directly harm a patient. Every procedure, deviation, and system must be documented and defensible.
A deviation is any departure from an approved procedure or specification. It could be minor — a temperature briefly outside range — or significant. What matters in GMP is not that deviations never happen (they will), but that they're detected, documented, investigated, and resolved properly.
When a deviation is identified, it triggers a formal investigation to understand root cause and assess product impact. If a systemic issue is found, a CAPA (Corrective and Preventive Action) is opened to fix the underlying problem. All of this is documented and available for regulatory inspection. The paper trail is the point.
Data integrity means every piece of data is attributable (you know who recorded it), legible, contemporaneous (recorded at the time it happened), original, and accurate. The FDA uses the acronym ALCOA to describe these principles.
In practice: electronic systems need tamper-proof audit trails, paper records must be completed in real time, and any correction must be documented with a reason. When I build operational tools, data integrity requirements shape every design decision — how data is captured, stored, timestamped, and retrieved. It's not an afterthought. It's the foundation.
This is something I've navigated personally. My background is in Management Information Systems and Marketing — not biology or chemistry. What opened doors was demonstrating I understood how manufacturing operations work as systems, and that I could build tools that made those systems run better.
A few things that genuinely help: understanding GMP concepts even without lab experience, being able to work with operational data, showing you've built something practical. Cell therapy facilities increasingly need people who sit at the intersection of operations and data — and that's not a pure science role.
More than people expect. MIS teaches you to think about organizations as information systems — how data flows, where it gets lost, where decisions happen and why. That thinking directly applies to manufacturing environments where data integrity, workflow design, and system architecture are constant challenges.
Where it's even stronger: combine MIS with domain knowledge. Understanding GMP requirements, knowing how manufacturing operations work, and being able to build tools makes you genuinely rare. The gap between people who understand the data and people who understand the operations is where a lot of real problems live — and where real value gets created.
From where I sit, these are the most valuable:
In a cryogenic storage environment, tanks hold samples in numbered racks, each with a set of locator slots. When placing a new sample, you need to know which slots are empty. The only source of truth was a daily EBS export — a flat file with thousands of rows — that had to be manually interpreted to find available locations.
That process was slow and error-prone, especially under time pressure. So I built a system that ingests the daily export, computes availability at the rack and slot level, and presents results as a filterable, color-coded dashboard. What used to take minutes of manual work now takes seconds. Built with Python (FastAPI), Next.js, and PostgreSQL, deployed with Docker.
The publication examines how artificial intelligence can be responsibly applied across cell therapy manufacturing — covering process optimization, quality governance, data architecture, and risk management. It's grounded in FDA guidance and the ICH Q10 quality system framework, so it speaks to AI not just technically, but within a regulated compliance context.
It's written for manufacturing scientists, quality leads, operations managers, and data professionals working in or entering the cell therapy space. The goal was a practical framework, not an abstract overview. You can read it here.
Absolutely — the best ways are email or LinkedIn, both listed on the contact section of the main page. I'm open to conversations about roles, collaborations, or questions about the field that aren't covered here.
If you have a question you'd like to see added to this hub, send it over. If it's something others would benefit from, I'll add it. This is meant to be a living resource.
If something isn't covered here, reach out directly. Happy to answer questions about cell therapy, data systems, or finding your path in this field.
Get in touch