Gil Raitses

Gil Raitses

I build software for research, from data pipelines and visual workspaces to the agent workflows that keep long, multi-repository projects coherent.

Research focus

I build computational infrastructure for multi-scale analysis across behavioral and sensor-driven systems, with an emphasis on representational coherence under variation in scale, resolution and structure.

The work blends neuroethology, signal processing and simulation with a pragmatic goal of keeping interpretability as internal descriptions shift.

Implementation landscape

Active themes I am implementing across projects form a working map of what I build and test today, while a retrospective summary would pretend that inventory is already settled.

Systems and behavioral neuroscience

  • Sensory-motor integration and behavioral outputs
  • Navigation and spatial cognition in three-dimensional environments
  • Attention, learning and memory in adaptive behavior
  • Cross-species behavioral modeling and trajectory analysis

Theoretical neuroscience

  • Representational geometry and predictive spatial maps
  • Social perception

Machine learning and methods

  • Adaptive windowing and feature extraction from complex signal data
  • Memory-efficient recurrent architectures for time series classification
  • Dynamic state-space modeling for behavioral transitions
  • Multimodal representation of continuous signal patterns

Framework implementations

aimez.ai

Applied research program that takes public traffic-camera frames through computer-vision models into a camera-derived crowd pressure field and into pedestrian routing that respects lived burden. It carries the work from raw field data to a public site with notes, figures and validation briefs.

wavves

Charter-driven orchestration keeps long multi-repository agent work coherent across sessions. Wavesets dispatch bounded workers under a fresh orchestrator while checkpoints and role contracts keep the audit trail local and inspectable.

Current work

2024 to 2026

Projects

Agent Orchestration2026

wavves

Charter-driven orchestration and audit keep long multi-repository agent work coherent across sessions.

Explore
Behavioral Detection2025

retrovibez

Larval reverse-crawl detection adopted by a mechanosensation group. Klein et al. 2015 method.

Explore
Data Pipeline2025

magatfairy

One-command bridge from legacy MATLAB analyzer data to HDF5 and Python, with validation and a CLI.

Explore
Spatial Computing2026

Klosr

City-scale evidence model for comfort-aware pedestrian routing on Manhattan. Sparse observations fuse onto a walk graph before route costs combine. Field-JEPA forecasts the next half-hour crowding field.

Explore
Marine Biology2024

whale behavior analysis

Interactive biologging visualization for dive telemetry, built to raise labeling and review throughput for feeding-behavior research.

Explore
Behavioral Simulation2025

IndySim

Interface dynamics simulation model and analysis code for Drosophila mechanosensory behavior, paired with the campus research manuscripts below.

Explore
Physical AI2026

orcast

Evidence-gated encounter forecasting for Southern Resident killer whales, with a biologging workbench and public 3D surface. Public launch this summer. Orca Strike is the live module below.

Play Orca Strike
Interactive2026Live

Orca Strike

Playable replay and reconstruction module for orcast on Bash. Interactive orca console in React and React Three Fiber.

Explore

Research program gallery

These cards preview projects from the broader research program, while each one opens a deeper page where hovering reveals the short description.

aimez

aimez builds evidence models and tools for decisions about complex physical systems. The current product focus is Klosr, a city-scale evidence model of Manhattan that turns sparse observations into decision-ready pedestrian representations, with comfort-aware routing as one application on that surface.

aimez.ai  ·  Klosr  ·  product overview  ·  repo

Research papers

Internal State and Extinction Durability in Retrieval Control

Gil Raitses

Septal GABAergic Timing in Entorhinal-Hippocampal Retrieval

Gil Raitses

On Decision-Valued Maps and Representational Dependence

Gil Raitses

Quantitative Guidelines for Behavioral Phenotyping from Sparse Point-Process Data

Gil Raitses, Devindi Goonawardhana, Mirna Mihovilovic Skanata

Temporal Dynamics of Mechanosensory Behavior in Drosophila Larvae

Gil Raitses, Devindi Goonawardhana, Mirna Mihovilovic Skanata

Mixed-Initiative Orchestration of a Managed-Agent Catalog

Gil Raitses · aimez.ai · 2026

Intent mining and provenance-gated narration in a human-AI chat console extend Horvitz's mixed-initiative principles to an orchestrated catalog of managed agents, with routing as the demonstration domain for the whole system.

Agentic Roles for Applied Physical-Intelligence Research

Gil Raitses · aimez.ai · 2026

A field-study platform and orchestrated routing act as a world-model grounding instrument, using Ha and Schmidhuber as an interpretive lens that grounds a build-time world model against field observation.

Physical-World-Model-Driven Adaptive Routing with Orchestrated Agent Mediation

Gil Raitses · 2026

Five comfort routing cost encodings run on the Manhattan island walk graph, where comfort and adaptive routes use per-bin MRT from a precomputed SOLWEIG model. The remaining gap to live weather data is documented with closure paths, while the product name for the live surface is Klosr. The detailed PDF remains withheld under the public IP boundary; the live product surface is on aimez.ai.

orcast, Evidence-Bounded Encounter Forecasting

Gil Raitses · aimez.ai · 2026

Evidence-gated encounter forecasting for Southern Resident killer whales in the Salish Sea keeps displayed confidence tied to held-out skill through validation gates. A biologging workbench and public 3D surface prepare a wider launch this summer, while Orca Strike remains the live public module on Bash.

Decoupling Mechanical and Biochemical Control of Gastruloid Symmetry Breaking

Gil Raitses, Daniel Podzunas, Xiangyu Zou, Eden Vallari, Jennifer Kang · NSF EmIRGE-Bio NRT · 2026

PHY 600 whitepaper proposing a falsifiable perturbation design that separates surface-tension driven tissue flow from morphogen patterning during organoid symmetry breaking.

Spring Network Allostery on an Erdős–Rényi Architecture

Gil Raitses · NSF EmIRGE-Bio NRT · 2026

PHY 600 report on multi-mechanism learning of spring constants so a random-graph network acquires a long-range allosteric input–output tension map on an ER baseline.

Formal Logical Analysis of AI-Generated Grant Reviews

Gil Raitses · NSF EmIRGE-Bio NRT · 2025

MAE 600 analysis quantifying structural and rhetorical patterns in AI-generated NSF CAREER grant reviews with propositional and decision-theoretic classification.

Technology and Responsibility Report: mRNA Technology

Gil Raitses (executive summary) with NRT teammates · NSF EmIRGE-Bio NRT · 2025

MAE 600 Technology and Responsibility team report on mRNA platform access disparities and dual-use risk. Gil authored the executive summary.

Get in touch

Researchers, collaborators and anyone interested in the orcast whitepapers, Klosr or future aimez.ai research releases can leave an email here. I will follow up directly with no newsletter, just the papers when they are ready to share.