A senior systems engineer moving into bio-computing. Three lenses on one pivot: the careers, what's worth building, and the substrate itself.
One door in, then branch: the companies hiring, the roles they hire for, and where to meet them. A crypto/DeFi/HPC background covers most of it; the gap is biology fundamentals, a 3-6 month crash course.
The single entry into the career side. A scannable employer directory — frontier AI-bio, synbio foundries, bio-data platforms, clinical/diagnostics — with the European hubs folded in (Basel, Cambridge, Mainz, Barcelona, Amsterdam, Prague), each cross-linked to the Positions Catalogue (the roles they hire for) and Conferences (where to meet them).
The apply-today list: current biotech / bio-SWE engineering openings paying $160k+ total comp, each with a direct apply link fetched and confirmed live. Split into salary-stated-in-posting (Profluent, Xaira, Freenome, Pfizer, Ginkgo, Recursion…) and frontier labs where pay is hidden but credibly clears the bar (Chai Discovery, Isomorphic, Insitro, Iambic). Closed and sub-threshold roles are called out, not padded in.
Biotech / techbio VCs that build companies in-house and hire engineers into the labs arm — not just scientists. EU-reachable leads (Deep Science Ventures, Sofinnova MD Start, Oxford Science Enterprises, Cambridge Innovation Capital) and the US relocation tier (Foresite Labs, 8VC Build). Each with a verified careers/EIR link and how an outside engineer actually gets in.
The source list monitored every week to catch founding-engineer / early-SWE / data-plane roles at fresh biotech spinouts and NewCos — the pragmatic backdoor (funds drop engineers into portfolio companies, not the fund). Funding outlets flag a launch → check the backer's portfolio + the NewCo's careers → target the first-engineer slot. Feeds the Openings + Opportunity + VC-Labs pages.
Standalone systems artifacts that earn the seat — solo-buildable, intrinsic value on their own, no AI-agent glue — plus the open places to search. The portfolio piece is the proof.
Frontrunner: a Rust QUBO/Ising solver (simulated-bifurcation, SIMD sparse matrix-vector inner loop). Quant/DeFi users: portfolio-under-cardinality, DEX arb-cycle detection. FOSS today is torch-Python; the production option, Toshiba SQBM+, is paywalled and is itself an Ising machine. Runner-ups: a zfp/SZ3 error-bounded float compressor (Rust has only FFI), a pure-Rust sparse direct solver (Cholesky/LDLt then LU; no production Rust exists), bgzf-turbo (parallel BGZF, doubling as a noodles contribution), and a deterministic fixed-point DSP.
Given a protein, simulate what it does in a cell — you can't, end to end. The five stages (structure → MD → the timescale wall → interactions → systems/cell), how each is handled and where it leaks, the full open-source landscape, and the recoverable gap: fast biological-timescale dynamics in Rust (no production Rust MD engine exists).
A no-fluff directory: UniProt, NCBI, Ensembl, PDB and AlphaFold DB, JASPAR and EPD (promoters & regulation), STRING and Reactome, plus the search engines (BLAST, MMseqs2, Foldseek, DIAMOND) and the query libraries (Biopython, gget). Notes where a Rust k-mer / alignment crate is the opening.
The wider field of substrates and models that compute without a von Neumann CPU. Context for why an Ising machine, a chemical reaction network, or a cell is a real computer, and where the systems-engineering leverage sits.
How the cell actually computes, how the field designs and engineers it, and what it would take to program it reliably. The mechanisms a builder needs before picking a wedge.
The old view (BioBricks, logic gates, abstraction) against the emerging one (foundation models, Evo 2, ESM3, the virtual cell). Where each wins, the iBiology series plus transcripts, and why "more like an LLM than a CPU" is the 2024-26 shift.
The Wendell Lim / UCSF cell-design school: synNotch receptors, engineered CARs, combinatorial AND-gate targeting, synthetic morphogenesis, engineered-cell therapies entering the clinic. Plus the ML-circuit layer and where a systems engineer fits, the Ginkgo/Tierra biofoundry world.
Reprogram a yeast to run a plant's chemistry and ferment the molecule: vanillin, Reb M stevia, valencene, ambroxide, squalane, HMOs are real products; saffron, menthol, cannabinoids, most pigments are research or dead. The commercial-vs-hype map, the methods (chassis, CRISPR, DBTL, the EU 4-year reg wall), EU employers, and the sharpest systems-engineer opening — a fast FBA/MILP solver in Rust, the same shape as the what-to-build thesis.
The field map (5 modalities), the 2026 state of the art (binders, antibodies, genome generation, what works vs hype), the companies hiring with EU/UK options (Cradle, Latent, Basecamp, Isomorphic), the systems-engineer wedge ("limited by engineering and compute"), and the competency-products to build.
The compiler that exists (Verilog to DNA at 60-75% first-pass), the de-facto toolchain, why the abstraction leaks (a gate is not a transistor), and the claim that this is ASIC tapeout, not gcc. Where a systems engineer has leverage.
Best-in-class simulators, the "data not FLOPs" wall (why a faster solver alone doesn't move the needle), and where speed actually pays: stochastic gene-circuit modeling and low-latency trajectory work. The wedge that maps to the build lens.
Where control theory meets the cell: Khammash's feedback control (reliability as a structural guarantee), the reprogrammable substrates, and the orthogonal wing (Levin bioelectricity, organoid intelligence, xenobots).
Last updated: July 2026 | krons.fiu.wtf