Executive Summary Running Qwen3.5-35B GPTQ Int4 on 4× Nvidia T4 16GB GPUs is feasible with vLLM through tensor parallelism, distributing model computation across all GPUs. The Qwen3.5-35B model (35B total parameters with 3B activated via MoE) has an estimated GPTQ Int4 footprint of approximately 8-10 GB, which requires tensor parallelism across all 4 GPUs (totaling 64GB) to achieve optimal performance. vLLM’s architecture, built on PagedAttention for efficient memory management and GPTQ quantization support, enables this configuration to deliver reasonable throughput for inference workloads while staying within T4 GPU memory constraints. However, performance will be substantially lower than on higher-end GPUs due to T4’s limited PCIe bandwidth (16× Gen3) and lower FP32 compute capability. ...

Introduction: The Closed Loop of Discovery For centuries, the scientific method has followed a familiar rhythm: a human scientist observes a phenomenon, formulates a hypothesis, designs an experiment, executes it manually or with basic automation, analyses the results, and iterates. This cycle — hypothesis, experiment, analysis, refinement — is the engine of scientific progress. But it’s also a bottleneck. Each iteration takes days, weeks, or months. Human bandwidth limits the search space we can explore. And crucially, the loop is open: the scientist must close it manually, bringing their intuition and experience to bear at every step. ...

The Dual-Use Dilemma In July 2024, the Arc Institute published a paper in Science describing Evo, a 7.6 billion parameter foundation model trained on 300 billion nucleotides spanning all domains of life. The model could generate functional DNA sequences, predict fitness effects of mutations, and even design novel regulatory elements. It was a scientific breakthrough—and immediately raised a question that every researcher in biological AI now confronts: Could this same technology be used to create biological weapons? ...

Introduction: The Open Science Paradox In May 2024, Google DeepMind published AlphaFold 3 in Nature, describing a system that could predict the structure of protein complexes with DNA, RNA, ligands, and small molecules—a dramatic leap beyond AlphaFold 2’s protein-only predictions. But there was a catch: the code wasn’t released. For six months, researchers could read about the breakthrough but couldn’t reproduce it, build on it, or verify the claims independently. ...

The Agentic AI space is maturing fast. This week brought clear winners in the framework wars, a convergence among coding agents, and a decisive shift toward enterprise security. Here’s what you need to know. The Multi-Agent Framework Landscape: Winners Emerge LangGraph: The Production Choice If you’re building agents that need to run reliably in production, LangGraph has become the default choice. Companies like Uber, LinkedIn, and Klarna have had LangGraph agents running in production for over a year. ...

Introduction: The Promise and the Peril Precision medicine promised to treat each patient as an individual — to move beyond one-size-fits-all therapies to interventions tailored to your unique biology. AI-driven omics seemed poised to accelerate this vision: algorithms that could read your genome, interpret your proteome, and predict your disease risk with unprecedented accuracy. But there’s a problem. The data powering these algorithms is profoundly unrepresentative of human diversity. As of 2024, over 94% of participants in genome-wide association studies (GWAS) are of European ancestry, despite Europeans comprising only about 16% of the global population. This imbalance isn’t just a statistical curiosity — it has real consequences. Polygenic risk scores trained on European data perform significantly worse for individuals of African, Asian, Hispanic, and Indigenous ancestry. Variant classification algorithms misclassify pathogenic mutations in underrepresented populations. And the AI tools now entering clinical practice risk cementing these disparities into healthcare systems worldwide. ...

The Ultimate Test: Does It Help Patients? After eighteen posts exploring the technical landscape of agentic omics—from foundation models for DNA and proteins to multi-agent systems for drug discovery—we arrive at the question that matters most: does any of this actually improve patient outcomes? The answer is more nuanced than the hype suggests. As of early 2026, the FDA has approved over 1,000 AI/ML-enabled medical devices, but only a small fraction operate on genomic or pathology data with demonstrated clinical utility (IntuitionLabs, 2025; Nature Digital Medicine, 2025). The gap between a model that achieves 95% accuracy on a benchmark and a tool that measurably extends survival remains wide—and crossing it requires navigating regulatory pathways, clinical validation studies, and the messy reality of healthcare IT infrastructure. ...

Introduction No single AI agent can master all of biology. A genomics specialist doesn’t reason like a proteomics expert. A literature review agent has different skills from an experimental design agent. Yet biological discovery demands all of these perspectives working together. This is the promise of multi-agent systems for biology: collaborative AI teams where specialized agents debate, coordinate, and peer-review each other’s work — mimicking the collaborative nature of real scientific teams. ...

Introduction: The Precision Oncology Imperative Cancer is not one disease but hundreds—each with distinct molecular drivers, treatment responses, and clinical trajectories. The promise of precision oncology is simple in concept but staggering in execution: match the right treatment to the right patient at the right time, guided by the molecular profile of their tumor. In practice, this requires orchestrating a complex workflow: tumor sequencing to identify mutations, interpretation of those variants against clinical databases, integration of genomic data with transcriptomic and proteomic profiles, therapy matching against drug databases, clinical trial matching, and longitudinal monitoring for resistance and recurrence. Each step generates data, requires expert interpretation, and carries uncertainty. ...