Accelerating Hyperspectral Biology with AI Agents. To build domain-specialized AI agents that can carry a hyperspectral analysis end to end: defining the task, proposing several algorithms, writing and running the code, comparing results, and then auditing the evidence for errors such as data leakage or failure on new conditions. They will evaluate the agents on public hyperspectral reporter data across varying backgrounds, light levels, and imaging distances.

Hyperspectral Characterization of Carotenoid- and Retinal-Producing Halophiles. To measure carotenoid- and retinal-producing halophiles with hyperspectral cameras. Salt-loving microbes produce pigments such as bacterioruberin and bacteriorhodopsin, which may act as optical fingerprints detectable from the air. He will fly a calibrated drone-mounted VNIR system over 30 sites in the Great Salt Lake's North Arm, pairing the imagery with shotgun metagenomics and HPLC pigment measurements. The lake has lost roughly two-thirds of its surface area since 1850.

HyperMix: open-source detection of engineered biosignatures in remote hyperspectral imagery. Engineered bacteria have been spotted from 90 meters away, but the signal is faint and hides inside the colors of soil, leaves, and water. He will build a physics-based scene simulator with known ground truth, reproduce the classical baselines, and train a physics-informed network that detects and unmixes against unknown backgrounds. MIT-licensed, with a public benchmark.

A foundational open reference dataset of epithelial tissue states using label-free hyperspectral imaging. No shared open reference currently links spectral signatures to independently verified states in living tissue. Working in the Drosophila wing epithelium, she will image healthy, wounded, and apoptotic tissue under matched ex vivo conditions, then use fluorescent markers to confirm each state so that every spectrum is anchored in known biology.

Multiplexed microbial whole-cell biosensing of environmental pollutants using hyperspectral reporters. Ecosystems are usually exposed to mixtures of contaminants rather than single compounds, but too few compatible reporters exist for multiplexed sensing. They will engineer predicted hyperspectral reporters in Saccharomyces cerevisiae and Pseudomonas putida, validate their actual spectra, and integrate them pairwise to determine which reporter pairs can be told apart.

How can a cheap scanner reveal hidden signals in living soil? Before a biological signal can be separated from its background, someone has to measure the background well. He will build a benchtop scanner from an off-the-shelf mini-spectrometer and a light-controlled chamber, then test whether it yields repeatable soil reflectance curves across moisture, texture, and organic matter gradients. Spectra, code, bill of materials, and build guide all released.

Spectral fingerprints of soil biology: hyperspectral imaging of fungi and plant roots in the soil environment. Fungal hyphae are only a few microns wide and can resemble roots or organic matter in visible-light images. Using reflected-light hyperspectral microscopy at UCSB, she will scan soil aggregates from coastal temperate rainforest sites to test whether hyphae, roots, minerals, and organic material are spectrally separable, and which wavelengths carry the most information.

A Hyperspectral “Colour Cloud” Framework for Detecting Environmental Disruption of Avian Visual Signals. Millions of specimens already record environmental conditions, such as historical black carbon pollution, within their plumage. He will image them at pixel-level resolution and develop an unsupervised “colour cloud” clustering algorithm to localize anomalous colour variation without a priori bias. Natural history collections as a massive distributed sensor network, extending monitoring backward in time.

A Spectral Atlas of Nature’s Metabolites: Reading the light-absorbance spectra of life, one molecule at a time. Absorbance spectra are recorded constantly, since most LC runs pass eluent through a diode-array detector, and then discarded. The bottleneck is not instrumentation; nobody has run it systematically across chemical space. Using the UCSB foundry's copy of the NCI natural-products library, they will validate a pipeline on authentic standards and publish every structure-linked spectrum openly, with full acquisition metadata.

March 2026

Laser-based PCR using infrared light and gold-coated tubes.

March 2026

Cell-free protein synthesis on a gel filtration column for higher yields.

March 2026

Same-day colony-to-sequence cloning using homebrew SPRI and nanopore.

March 2026

Recipient requested anonymity.

March 2026

An engineered ribosome that translates protein directly from DNA.

March 2026

A centrifuge-based liquid handler, a DNA-origami protein printer, and a label-free logic analyzer for nanoscale devices.

March 2026

Evolving Vibrio for sub-10-minute doubling and pooled survival assays for protein binders.

March 2026

Antifreeze-protein-driven freeze concentration to accelerate ligation, and designed cradles for membrane protein expression.

March 2026

An automated cloning designer and a universal donor strain for multi-plasmid conjugation.

March 2026

Turning the giant single-celled alga Valonia ventricosa into an injectable biotech chassis.

March 2026

An open-source benchtop system that automates colony PCR and gel electrophoresis end-to-end.

March 2026

Designing oligo pools so a single PCA reaction yields a custom protein variant library.

March 2026

Experiment Engine: an overhead camera that auto-logs every step of bench experiments.

March 2026

IceCreamClone: an interactive tool that ranks DNA assembly strategies and teaches cloning logic.

March 2026

Recipient requested anonymity.