Catalog · 06 entries / Last updated · MAY 2026 / Filter · ALL DOMAINS

Selected work across signal, vision, and AI.

A representative slice of public and prototype work. Many engagements are under NDA and are not listed here — ask us about adjacent capabilities.

FEATURED · 01

Multi-agent sonar tracker

Cooperative underwater target tracking across a distributed passive-sonar sensor network. Open-source reference implementation.

● OPEN SOURCE SONAR MULTI-AGENT TRACKING

multiagent-sonar-tracker

N cooperating agents observe a moving underwater target through passive bearing-only sonar. Each agent maintains a local state estimate and fuses information with neighbors to converge on a shared track. The simulation exposes the geometry, motion-model, and communication trade-offs that drive performance in real distributed sensor networks.

Python, NumPy, matplotlib. MIT-licensed. Reproducible on commodity hardware.

View on GitHub ↗ Request briefing

DomainUnderwater acoustics

MethodDistributed estimation, data association

SensorsPassive sonar, bearing-only

AgentsN ≥ 2, configurable

StackPython · NumPy · matplotlib

LicenseMIT

Status● Active

CATALOG · 02–06

Prototype catalog

Adjacent capabilities we have built or stood up in research settings. Detailed write-ups available on request.

02 / RADAR

SAR target classification

Deep-learning classifier for synthetic aperture radar imagery under low-look-angle and clutter-limited conditions. Internal prototype, MSTAR-style benchmarks.

03 / VISION

Low-SNR object detector

Detection pipeline that holds precision as image SNR degrades — augmentation, denoising front-end, and uncertainty-aware decision head.

04 / 3D

Sparse-view 3D reconstruction

Neural-field reconstruction from a handful of viewpoints. Targets surveillance and rapid-mapping use cases where dense capture is impractical.

05 / GEN-AI

Synthetic data augmentation

Diffusion-based synthesis of rare-class samples for downstream detectors. Closes the long tail when real collection is expensive or restricted.

06 / DSP

Adaptive beamforming sandbox

MVDR, GSC, and LMS adaptive arrays with a swappable interference model. Built for teaching, baselining, and rapid algorithm comparison.

07 / TBD

Your problem here

Engagements typically start with a 1-page scoping note. Start a conversation.

Engagement

Want a deeper look at any of these?

Each prototype has a private write-up with architecture, data sources, and benchmark numbers. Send a note and we’ll share what we can.

Email[email protected]

GitHubalexberian/multiagent-sonar-tracker

ScholarA. Berian — publications