Make the first
scientific step
easy to take.

# Week 00 — personal baseline
Status: DRAFT — fill with Peter's own answers.
Date:
Preferred public name:
Current role / organization (only what can be public):
Career direction: applied scientific AI in biotechnology and pharma.
Weekly availability / exact calendar blocks:
Location, languages, remote/travel preference:
Work authorization constraints if relevant:
Role route to test: applied scientist / scientific AI lead / both.

## Skill evidence
| Capability | Can do alone / with help / new | Evidence link or example | First gap |
|---|---|---|---|
| Python and debugging | | | |
| Git and reproducible environments | | | |
| Statistics and evaluation | | | |
| ML / LLM systems | | | |
| Biology and bioprocess vocabulary | | | |
| Leading delivery and teams | | | |

## Three defensible career stories
For each: context, my responsibility, action, result, supporting evidence, what is confidential.
1.
2.
3.

## Five physical next actions
1. Open cockpit/TODAY.md and name Monday's next action.
2. Fill this baseline without inflating unfamiliar skills.
3. Open the v5 master at Part 0 and explain the working rhythm.
4. Create the three NotebookLM notebooks and check one citation.
5. Run the Python environment smoke check and save its output.

## Budget and uncertainty
API spending authorized:
Equipment actually available:
Claims still needing verification:

# Week 00 — orientation and learning setup
Status: NOT_STARTED. No biology course completion is required this week.

## Source order
v5 governs; v4 supplies retained offers/company map/templates; weekly specs expand execution.
Read v5 Part 0 twice and Part 14 once; inspect Parts 3 and 5.

## My five operating rules (write in my own words)
1.
2.
3.
4.
5.

## Notebook records
| Notebook | Link | Uploaded sources / version | Citation checked |
|---|---|---|---|
| LABLOOP — Bio | | | |
| LABLOOP — ML | | | |
| LABLOOP — Thesis | | | |

## My teach-back
Why am I building an instrument and a dataset?
What can a reproducible result prove? What cannot a DOI prove?
What is the next physical action if I get stuck?

## Quiz (my answers before reading the answer key)
1. What is the difference between TODAY and PROGRESS?
2. What is my smallest useful Week 00 outcome?
3. Which source wins when v4 and v5 disagree?
4. Does a submitted manuscript mean a public/peer-reviewed paper?
5. If I lose a week, what is the recovery action?

Actual score / feedback:
Unresolved questions:

# batchlens
Early-stage research tooling for fermentation data.

Status: Week 00 skeleton. No scientific performance result or release DOI yet.

## Question
How can we make fermentation observations, deviations and predictions reproducible and inspectable?

## Available now
A synthetic environment smoke check, once scripts/smoke.py is saved and run.

## Smoke check
    .\.venv\Scripts\python.exe scripts\smoke.py --fake

## Data
No plant or client data. The smoke output is synthetic and has no biological meaning.
Public benchmark manifests and checksums will be added in the relevant week.

## Scientific results
Not yet measured. Report actual sample sizes, independent splits, uncertainty and limitations later.

## Reproduction
Only the smoke check is available at setup. make report and make paper are future interfaces, not working commands today.

## Limitations
No validated process model, clinical use, or production monitoring claim.

## Citation
CITATION.cff records project identity. No DOI has been issued.

# labloop-notes
Selected study notes and build logs for LABLOOP.

Status: Week 00 skeleton.
Folders: notes/bio/, notes/ml/, notes/papers/, logs/.
Review material before publishing. Private career information and company-gap hypotheses stay in cockpit or the planning folder.
No course completion or scientific result is claimed.

# fermlit
A planned reproducible workflow for curating fermentation-literature data.

Status: Week 00 skeleton. No dataset, accuracy score or preprint yet.
Question: Which experiment details can be extracted with provenance and verified against human annotations?
Planned structure: corpus/, schema/, annotate/, extract/, eval/, analysis/, dataset/, paper/.
No papers or client data are redistributed at setup.
Quickstart and measured evaluation will be added when implemented.
No release DOI yet.

.venv/
__pycache__/
*.pyc
.env
.env.*
!.env.example
data/raw/
private/
.DS_Store

cff-version: 1.2.0
message: "Development project; no scientific result or release DOI is claimed."
title: "batchlens"
type: software
authors:
  - name: "REPLACE WITH VERIFIED PUBLIC NAME"
# Change title to fermlit when copying to that repository.
# Add repository URL only when it exists; omit DOI until one is issued.

# Week 00 — environment and repository evidence
Status: NOT_STARTED. Record actual outputs; commands in the guide are instructions.
Date:
Local repository root:
Canonical private cockpit path:

| Item | Result / version / local path | Check performed | Status / next action |
|---|---|---|---|
| Git | | git --version | NOT_STARTED |
| uv | | uv --version | NOT_STARTED |
| Python 3.11 | | python version from each .venv | NOT_STARTED |
| batchlens .venv | | math smoke example | NOT_STARTED |
| fermlit .venv | | Python version | NOT_STARTED |
| Fake pipeline | | smoke.py --fake prints synthetic JSON | NOT_STARTED |
| RAM / GPU / VRAM | | actual hardware inventory | NOT_STARTED |
| Ollama | | ollama --version | NOT_STARTED |
| Text model | | model name, response, ollama ps | NOT_STARTED |
| Vision model | | one synthetic image test | NOT_STARTED |
| Embedding model | | numeric vector returned locally | NOT_STARTED |
| CPU fallback | | Python fake path works without GPU | NOT_STARTED |
| Git ignore | | git check-ignore on .env | NOT_STARTED |
| README and CITATION.cff | | no invented results/DOI | NOT_STARTED |
| GitHub visibility | | cockpit private; public skeletons reviewed | NOT_STARTED |

## Repositories
| Name | Local path | Remote URL / not created | Visibility | Last commit |
|---|---|---|---|---|
| cockpit | | | PRIVATE | |
| labloop-notes | | | PUBLIC target | |
| batchlens | | | PUBLIC target | |
| fermlit | | | PUBLIC target | |

## Download or setup problems
Problem / observed message:
Smallest next check:
Owner / next date:
Data or secrets to keep out of public repositories:

# Week 00 — private relationship map
Status: RESEARCH SEED. No messages sent.
Use v4 sections 3.1–3.3 and Appendix A plus GridX Startups.csv.
Company gaps are hypotheses. Verify official pages and current roles before contact.

| Track | Company/person candidate | Source / checked date | Why relevant | Proof needed before approach | Next action |
|---|---|---|---|---|---|
| Scientist / feedback partner | | | | | |
| Employer | | | | | |
| Recruiter / talent contact | | | | | |
| Future client | | | | | |

Select five initial candidates across these tracks; no active vacancy is assumed.
Examples to investigate from the plan: Michroma/Levya for fermentation; DataHow/Aizon for scientific-data roles.
Recruiter entry: identify a named person from a current official or professional profile rather than invent one.

## First relationship step
Read paper -> useful question -> relevant finding -> conversation -> feedback -> collaboration -> qualified offer (later in v5).
Week 00 action is research and preparation.

## Six culture questions from v4
Who decides priorities?
What happened when a milestone slipped?
How long is an experimental cycle?
Is there on-call, and for what?
How many people left recently, and why?
Can we discuss recent retrospectives or lab meetings?

Selected next paper:
Next verification action:

# Week 00 — candidate paper shortlist
Status: SHORTLIST ONLY. The recurring full digest begins in Week 01.
Select one candidate connected to fermentation, e.g. a verified Michroma or Levya scientific author from the source company map.

| Candidate | Official company/lab page | Paper title / DOI / URL | Author connection verified? | Full text available? |
|---|---|---|---|---|
| | | | | |

Why this paper connects to the first simulator:
One question after checking the abstract/methods:
Paper chosen for Week 01:
Do not infer methods from a title or mark a paper read before reading it.

# Week 00 — positioning
Status: DRAFT. Verify personal claims and paste final published links only after publication.

## Headline
CEO & applied AI practitioner | Building toward scientific AI in biotech and pharma | Reproducible tools for fermentation and lab data

## About
I am bringing my experience in business leadership and applied AI into biotechnology and pharma.

Through LABLOOP, I am developing practical scientific skills by building reproducible tools and datasets for fermentation and laboratory data. My first focus is measurement: how to turn experimental records into data and models that other people can inspect.

I will document the methods, results and limitations as the work progresses. I am interested in conversations with scientists, research engineers and companies working at the intersection of AI and experimental R&D.

## Applied-scientist route
I am building hands-on evidence in scientific data curation, evaluation and bioprocess modeling. I am interested in roles where I can contribute my applied-AI experience while deepening my scientific practice.

## Scientific-AI-lead route
I connect business priorities, technical execution and measurable delivery. My focus is developing the scientific depth to lead useful, reproducible AI work with biotech and pharma R&D teams.

## GitHub bio
Applied AI for biotech & pharma. Building LABLOOP: reproducible fermentation and laboratory-data tools. Learning and results in public.

## 30-second introduction
I am a CEO and applied-AI practitioner developing a scientific-AI focus in biotechnology and pharma. I am starting with fermentation and laboratory data: reproducible tools, evaluated models and transparent results. I bring leadership experience, and I am building the scientific evidence step by step. I am looking to learn from people working close to experiments.

## Claim audit
| Claim | Evidence | Public permission / correction needed |
|---|---|---|
| Current CEO role | | |
| Applied-AI experience | | |
| Any years, employer name or quantified outcome added later | | |

Final LinkedIn URL:
GitHub profile URL:
ORCID URL:

# Week 00 — announcement drafts
Status: DRAFT — adapt to what was actually completed.

## Sunday LinkedIn post
Week 00 of LABLOOP: a 40-week learning and building project focused on scientific AI for biotechnology and pharma.

I am starting with fermentation and laboratory data. The aim is to build tools and datasets whose methods, results and limitations can be inspected—not just demos that look convincing.

This week I [insert the setup actions actually completed]. My starting strengths are [verified strengths]. The scientific areas I need to learn are [honest gaps].

The first technical step is a small, clearly labeled fermentation simulation. It will be a learning model, not evidence of performance in a real plant.

I plan to share what works, what fails, and the evidence behind each conclusion. If you work with fermentation or experimental data, I would value a perspective on what makes a result useful to you.

[Include only an existing public link.]

## Optional Wednesday note
Setting up a scientific-AI project starts with a practical question: what evidence would let someone else check my work? I am organizing the environment, source notes and reproducibility checks for LABLOOP. The first question I want to investigate is [your own question]. No results yet; this is the starting point.

## Short X adaptation
Week 00 of LABLOOP: building scientific-AI skills for biotech and pharma through reproducible work. Starting with fermentation and lab data. Setup: [one completed action]. Next: a small simulation with explicit assumptions. [Actual public link, if available]

## Figure brief
Screenshot the four repository names and their honest README status.
Exclude private cockpit content and account details.
Caption: LABLOOP setup, Week 00. Scientific results are not yet available.
Alt text: Three public project skeletons for notes, fermentation tools and literature data.

## Publication record
Draft verified against evidence:
Published date / URL:
If still draft, why / next action:

# Week 01 — proposed physical actions
These are a handoff, not work to start during Week 00.
Mon: Open batchlens/sim/monod.py; write the variables, units and assumptions before implementing one balance.
Tue: Open the selected Biology 2e section; define protein, enzyme and substrate in my own words.
Wed: Open the telemetry runbook; identify the simulator sample format before configuring replay.
Thu: Open the selected linear-algebra lesson; work one vector/matrix example, then digest the chosen founder paper.
Fri: Open the simulator configuration; define seeded batches and one explicitly labeled fault.
Sat: Open the earliest incomplete gate-critical artifact and finish its smallest checkable step.
Sunday: build log, evidence check, gap review and the next six actions.
Move only the next physical action into cockpit/TODAY.md.