I think the bottleneck in modern biotech isn't the biology; it's the infrastructure underneath it. Most of my time goes into building the compute layer that sits between raw sequencing reads and something a clinician can actually act on.
What I find endlessly interesting right now is where multi-agent AI systems meet computational biology. Not as a buzzword, but as an engineering problem: can you build AI agents that reason over experimental context, triage QC failures, and orchestrate downstream analysis the way a senior bioinformatician would? I'm actively building toward that.
pipelines: Nextflow, Snakemake
languages: Python, R, SQL, Bash
cloud: AWS (S3, Batch, Lambda, Fargate, EventBridge)
infra: Docker, Terraform, Linux, HPC/Slurm
genomics: variant calling, cfDNA, ChIP-seq, RNA-seq, scRNA-seq, multi-omics
exploration: multi-agent orchestration, agentic workflows, LLM tooling- How to make genomics pipelines less fragile and more self-aware
- Whether AI agents can meaningfully reduce the operational burden in clinical labs
- The architecture that lets a team go from sequencing reads to insight without the pipeline becoming the bottleneck
- Why the best bioinformatics infrastructure looks a lot like good software engineering
Columbia MA Biotechnology '23 | IIT Kanpur BS Chemistry '19 | Published in Inorganic Chemistry