Build and own an AI-powered ingestion & normalization pipeline to import data from a wide variety of sources — unprocessed Excel/CSV uploads, lab and instrument exports, as well as processed data from internal pipelines.
Develop robust schema mapping, coercion, and conversion logic (think: units normalization, metadata standardization, variable-name harmonization, vendor-instrument quirks, plate-reader formats, reference-genome or annotation updates, batch-effect correction, etc.).
Use LLM-driven and classical data-engineering tools to structure “semi-structured” or messy tabular data — extracting metadata, inferring column roles/types, cleaning free-text headers, fixing inconsistencies, and preparing final clean datasets.
Ensure all transformations that should only happen once (normalization, coercion, batch-correction) execute during ingestion — so downstream analytics / the AI “Co-Scientist” always works with clean, canonical data.
Build validation, verification, and quality-control layers to catch ambiguous, inconsistent, or corrupt data before it enters the platform.
Collaborate with product teams, data science / bioinformatics colleagues, and infrastructure engineers to define and enforce data standards, and ensure pipeline outputs integrate cleanly into downstream analysis and storage systems.