Soroush Sarabi · FDE · Industrial & Scientific AI

Hire an FDE →

FDE roles · Forward-deployed engagements Canada · UK · EU · Turkey · UAE

With the team.
On their data.
In production.

Soroush Sarabi — Forward Deployed Engineer for applied AI in industry and science. I embed with engineers, scientists and clinicians, learn the problem on their own data — images, sensor signals, satellite and telescope data, documents — and ship the system into production, then stay until it holds. Machine vision, remote sensing and environmental monitoring, thermal design, clinical neuroscience, genomics and astronomy; LLM and document systems where a language model is the right tool. Behind that: 25 years of engineering, Managing Director of Raderon AI Lab, named inventor on six WIPO PCT international filings, peer-reviewed in MNRAS, PASP, IEEE GRSL and Thermal Science and Engineering Progress.

See the work → Read the research ↓ scroll

Soroush Sarabi — FDE · Industrial & Scientific AI

SS · 001 · v.2026.11

Identity SS-001 · v.2026.11

Role Forward Deployed Engineer · Industrial & Scientific AI

Ventures Raderon AI Lab · Green Box AI · BLZN.AI

Operating Canada · UK · EU · Turkey · UAE

Languages Persian · English

Doctrine Embedded · their data · until it ships

Standard résumé

Technical CV ↗ Academic CV ↗

Years in software

WIPO PCT filings · named inventor

Publications

Google Scholar

182

Citations

Active product lines

Operating entities

Google Scholar

h-index

Google Scholar

i10-index

/01 The Loop

Few individuals close the entire loop on their own.

embed → build → ship

Stage 01 Embed

On their own data.

I sit with the engineers, scientists and clinicians where the work happens: plant, lab, clinic or law office. The problem is learned on their own images, sensor signals, satellite scenes and records, not on a demo dataset.

Plant Lab Clinic Field data

Stage 02 Build

Research-grade methods, real problems.

Methods I have published or patented, applied to their problem: machine vision, remote sensing, thermal-design surrogates, MCMC fitting, clinical signal processing and, where it fits, LLMs. Peer-reviewed in MNRAS, PASP, IEEE GRSL and TSEP; six WIPO PCT filings behind them.

MNRAS PASP IEEE GRSL TSEP WIPO PCT

Stage 03 Ship

Into production.

SyneuraX clinical VR rehabilitation. Satellite air-quality, fire and vegetation monitoring. The heatsink-design software behind the TSEP paper. A law-firm system that drafts under lawyer review. Deployed code, not slides, and I stay until it holds.

WPF ASP.NET Core Unity / Quest SQL Server Azure

/02 Selected Work

Nine product lines, one principal.

six representative

6 selected of 23 projects View all work →

VR · Neurorehabilitation /01

SyneuraX

Brain and body rehabilitation in VR, steered live by the therapist.

SyneuraX turns clinic rehabilitation into measured exercise in virtual reality. Patients train on standalone Meta Quest headsets while the therapist runs every headset from one Windows station, watches the live view, adjusts the session on the spot and gets objective reports. Thirteen clinical exercises are released and seven more are built and awaiting release, all on the clinic's own network with no cloud dependency.

One therapist station runs many Meta Quest headsets over a zero-configuration LAN protocol
13 clinical exercises built on Corsi, WCST, Tower of London, Go/No-Go and Trail Making
.NET 10 WPF · Unity 6 · Kotlin · ASP.NET Core · 3,501 automated tests passing at release

.NET 10 WPF Unity 6 Meta Quest ASP.NET Core Kotlin

Details →

News Intelligence · AI Pipeline /02

Jam-e Jamshid

A system that reads the world's news, understands it, translates it and publishes it.

Jam-e Jamshid is an automated news system that carries a story from the moment it appears to its publication on the web and Telegram. It watches sources, captures pages in a real browser and passes the text through a chain of AI steps, each with its own main model and fallbacks. Alongside that pipeline runs a daily tension index built on open-source data.

A multi-step AI pipeline with a main model, ordered fallbacks and its own validation at every step
The full chain, from source monitoring and page capture to Persian translation and publishing on the web and Telegram
A daily tension index and Bayesian forecast with credible intervals, on open-source data

.NET 10 ASP.NET Core SignalR CefSharp SQL Server Telegram API LLM Pipelines

Details →

LegalTech · Law Firms /03

Law Firm System

AI drafting under lawyer review, from intake to a court-ready file.

An end-to-end platform for law firms. Clients open a case and upload documents; a per-service AI pipeline extracts the facts, finds the legal basis in a curated corpus and live research, drafts from lawyer-approved models, a critic agent reviews the draft, and a lawyer approves it on a shared review desk before the client sees it. Around it: case and deadline management, a client portal, billing and subscriptions, and a full audit trail. Deployed for Canadian Federal Court immigration filings (Mandamus, memoranda of argument) and for a civil-law court system.

Per-service AI pipeline: facts, legal basis, drafting, critic review, lawyer approval
Anti-hallucination guards: every date and amount traced to the inputs; client-statement and scope checks
Versioned, evaluated AI releases; staff MFA, audit log and encrypted provider keys

.NET 10 ASP.NET Core SQL Server Multi-LLM pipeline RAG QuestPDF

Details →

Thermal Engineering · CFD /04

HeatSink Optimization Suite

Parametric heatsink generation and pressure-drop analysis.

A multi-project tool suite for thermal-design research: a parametric heatsink generator backed by an EF6 design database, a pressure-drop calculator for dimpled-nub geometries, and a Dimple analysis module. The scientific software behind peer-reviewed work in Thermal Science & Engineering Progress (TSEP) on heatsink optimisation.

Parametric heatsink generator with EF6 design database
Pressure-drop (Δp) solver for dimpled-nub geometries
Backed peer-reviewed TSEP publication on heatsink optimisation

WPF Entity Framework 6 SQL Server CFD TSEP Thermal Engineering

Details →

Astrophysics · Scientific Software /05

Binary Star Network Analyzer

Contact-binary modelling with MCMC and light-curve analysis.

A WPF analyzer for contact-binary star systems used in the international BSN Project (R/AST/2021/1001). Ingests time-series photometric data (Catalina Sky Survey and others), fits binary parameters with Markov Chain Monte Carlo, models starspots and inclination, and renders Hertzsprung-Russell diagrams — the scientific software behind peer-reviewed publications in MNRAS and PASP.

Markov Chain Monte Carlo parameter estimation with KDE and contour plots
Contact-binary modelling — Hertzsprung-Russell diagrams, light curves, starspot fits
Catalina Sky Survey time-flux ingestion · WPF desktop · BSN Project (R/AST/2021/1001)

WPF MCMC Photometry Contact Binaries BSN Project Astrophysics

Details →

Applied AI · Multi-Model Platform /06

Blue Horizon AI

A Persian-first platform that puts the world's leading AI models behind one interface and one wallet.

Blue Horizon AI brings dozens of language, image, video and audio models under one Persian interface, with toman payment and a single USD-denominated wallet. Next to chat and creative studios it exposes an OpenAI-compatible API, so developers connect by changing a base URL. Its standout strength is the money engine: every request is held before it runs and settled after, and a built-in auditor re-checks the books every six hours.

Multi-provider model gateway behind one catalogue; free models appear under Persian names in the «افق آبی» tier.
USD wallet with atomic hold-then-settle billing, a reseller layer and a self-auditing financial engine.
.NET 10 Razor Pages web, a native Android app in Jetpack Compose, Telegram bots and an OpenAI-compatible /v1 API.

.NET 10 ASP.NET Core SQL Server Jetpack Compose OpenAI-compatible API AES-256-GCM

Details →

/03 Selected Publications

Peer-reviewed in MNRAS, PASP, IEEE GRSL.

2025

IEEE GRSL

Enhanced-HisSegNet · SAR Image Flood Segmentation with Learnable Histogram Layers and Active Contour Model

Asadi · Sarabi · Kordani · Asghari Ilani · Banad — 22, 1–5

—new

2025

ML w/ Apps

Enhancing Skin Cancer Diagnosis using Late DWT and New Swarm-Based Optimizers

Mousa · Chamani · Morsali · Kazzazi · Hatami · Sarabi — 23, 100811

—new

2023

Fundamental Challenges of Neural Network in Handling Sequential Feature of Time Series: NP-Hard Challenge

R Mousa, S Sarabi

0citations

2022

MNRAS

Investigation of orbital period & mass relations for W UMa-type contact systems

Poro · Sarabi · Zamanpour et al. — 510(4), 5315–5329

37citations

2022

PASP

Photometric Study of Six W UMa Systems & Mass–Radius Relations

Poro · Paki · Blackford · Davoudi · Aladag · Zamanpour · Sarabi — 134(1036), 064201

13citations

2022

TSEP

Machine learning-based optimization of air-cooled heat sinks

Shaeri · Sarabi · Randriambololona — 34, 101398

34citations

2022

ICFFHMT

ANN Models to Predict HTC and Pressure Drops in Cold Plates with Surface Roughness

Shaeri · Randriambololona · Sarabi

6citations

2022

MCM '22

Prediction Accuracy of ANNs in Thermal Management Subject to Network Architectures

Shaeri · Sarabi · Randriambololona · Shadlo

9citations

2022

Proceedings of the 9th International Conference on Fluid Flow, Heat and Mass …, 2022

Artificial neural network to predict pressure drops in heat sinks

BN Mengesha, MR Shaeri, S Sarabi

9citations

2022

Proceedings of the 8th World Congress on Mechanical, Chemical, and Material …, 2022

Application of machine learning to predict thermal performances of heat sinks

BN Mengesha, MR Shaeri, S Sarabi

6citations

2021

PASP

Observational and Theoretical Studies of 27 δ Scuti Stars · Period–Luminosity Relation

Poro · Paki · Mazhari · Sarabi et al. — 133(1026), 084201

Acknowledges Raderon Lab — contract R\AST\2021\1001.

33citations

2019

JOE-Journal of Occultation and Eclipse, 2019

OC Study of 545 Lunar Occultations from 13 Double Stars

A Poro, S Memarzadeh, A Halavati, M Pakravananfar, I Safaei, ...

2citations

2018

arXiv preprint arXiv:1809.06907, 2018

Evaluating the effect of four unknown parameters included in a latitudinal energy balance model on the habitability of exoplanets

M Bahraminasr, SJ Jafarzadeh, F Montazeri, A Poro, S Sarabi

3citations

/04 Intellectual Property

Six named-inventor WIPO PCT filings.

public prior art

WO 2020/217082 A1 /01