Kim Fehl (Adameyko)
banner
kimfehl.bsky.social
Kim Fehl (Adameyko)
@kimfehl.bsky.social
Bioinformatician, PhD in #Genetics with a strong math background | Studying regeneration in marine sponges | #OpenToWork for postdoc or industry roles
linkedin.com/in/kim-fehl
In these datasets, one can spot beautiful patterns, even appeal to other senses — as did two artists in the Art & Science session, turning metabolic heatmaps into real sponge techno. The challenge ahead is to turn big data not just into sounds — but into music, an integral story.

9/9
September 15, 2025 at 12:00 PM
So I felt a bit adrift among experimental biologists, yet inspired to shift the scale of thinking.

I see life not through a microscope but through data...

8/9
September 15, 2025 at 12:00 PM
At the closing dinner I asked her how one dares to make such bold hypotheses. She replied: read widely, and always look for the most interesting biological questions.

7/9
September 15, 2025 at 12:00 PM
Sally Leys, who has worked on sponges for more than two decades — long before any genomes — gave a fascinating talk on sponge larvae functioning like compound insect eyes, tracing a 19th-century idea into modern genetics.

6/9
September 15, 2025 at 12:00 PM
I’ve recently started building a pipeline to integrate these data and even included example output on my poster about heat-shock proteins. It fit the climate theme, but showing that sponges upregulate stress genes much like other animals was unsurprising for such a seasoned audience.

5/9
September 15, 2025 at 12:00 PM
However, my primary interest, genomics, was scarce. Despite the Aquatic Symbiosis Genomics project releasing many new assemblies, the community still needs time to absorb them.

4/9
September 15, 2025 at 12:00 PM
Both field observations and tank studies mimicking acidifying, warming seas gave hope: sponges may cope with climate change better than corals.

3/9
September 15, 2025 at 12:00 PM
For four days I filtered currents of knowledge: physiology, ecology, taxonomy— and some memorable experiments. Sponges glued to platforms and shifted between bays and depths; 3D scanning to track contractions and growth; tiny tubes inserted into pores to track respiration and bacterial ”menus“.

2/9
September 15, 2025 at 12:00 PM
The result: AGAT scripts are now pipe-friendly, console output is no longer chaotic, and tests, refactored from monolithic blocks, run about twice as fast. “It ain’t much, but it’s honest work.”

9/9
August 29, 2025 at 1:05 AM
Overall, it took longer than I thought — more than a week, with the last few days mostly spent manually debugging regression tests — making sure my additions wouldn’t break old behavior.

8/9
August 29, 2025 at 1:05 AM
Each coding task launches its own environment, so it’s worth teaching your agents to set it up efficiently via an AGENT.md instructions. Sessions last ~15 minutes, agents tend to cut corners near the end, so I recommend checking not only commits but also the endings of their “thinking” logs.

7/9
August 29, 2025 at 1:05 AM
You can even manage it from your phone (not good for sleep!). Soon I got used to breaking tasks apart for multiple agents working simultaneously so that merges weren’t painful.

6/9
August 29, 2025 at 1:05 AM
AGAT is not small either: almost a hundred scripts and library modules, spanning ~45k lines of Perl code. You really need a small coding team!

And if you have a ChatGPT Plus subscription, you already have one: Codex. Add GitHub Actions for testing, and your development is fully in the cloud!

5/9
August 29, 2025 at 1:05 AM
Last Christmas I took on too ambitious a project to learn how to use AI coding agents — trying to parallelize a large C tool by porting it to Rust with Cursor. (It was fun, though, to learn a new language by tinkering with software I use so often.)

4/9
August 29, 2025 at 1:05 AM
But when I tried to plug it into a pipeline, stray log messages leaked into the output stream. I logged an issue on GitHub. Simple hacks were possible, but I like elegance everywhere — my data flows included — so I realized it was a manageable task to contribute.

3/9
August 29, 2025 at 1:05 AM
As USB-C ended the zoo of power plugs, one day bioinformaticians won’t remember today’s zoo of text-based formats for genomic annotations. For now, AGAT (from @nbis.se) is one of the few toolkits designed to be proudly omnivorous.

2/9
August 29, 2025 at 1:05 AM
So I hope to finally find a new team where I can grow — and eventually, lead. Coming from a small, 'boutique' lab studying marine sponges,I'm now eager to contribute to broader #evodevo or #compgen projects. If you know of #postdoc openings in such groups, please reach out!

7/7
July 7, 2025 at 10:03 PM
Since leaving Russia in 2022, mentoring student projects has helped maintain a sense of connection. During thesis writing,I had to step back, and sadly, no new students joined last year. Meanwhile, AI has gotten smarter — but let's be honest, nothing compares to discussion with real colleagues.

6/7
July 7, 2025 at 10:03 PM
I was hoping for more inspiring interaction within a professional community, but ironically, our small department turned out to be fully remote, with everyone focused on individual tasks. Only shortly before I left did I convince our lead to introduce regular seminars and code reviews.

5/7
July 7, 2025 at 10:03 PM
This delayed feedback can be tough, especially when working remotely and largely alone. At one point, I grew so tired of being the sole bioinformatician in my lab — with collaborators distant and busy — that I moved into an industry role at Genotek.

4/7
July 7, 2025 at 10:03 PM
That first real citation. Not a self-reference in your next paper, but evidence that your work was genuinely seen, considered relevant, and integrated into someone else’s scientific reasoning — proof that it entered the slow blockchain of knowledge.

3/7
July 7, 2025 at 10:03 PM
Eventually, though, the results form into a coherent narrative. After a few more months of peer review, a neatly typeset PDF finally appears online — sometimes even in print. Yet the true milestone often arrives much later.

2/7
July 7, 2025 at 10:03 PM