Skip to content
Abstract neural vascular network visualization

Research Institute

Applied research across disciplines.

We publish open-source tools and peer-reviewed research. Every project ships code, data, and a paper. The domains grow with the work.

Mission

Build it. Open-source it. Publish it.

The STI Research Institute contributes original research and open-source tools across any discipline where applied science can push the field forward. No fixed lanes — the portfolio expands with every new question worth answering.

Reproducible science. Usable tools. No paywalls.

Research Programs

Active research tracks

Publications

The work speaks for itself

2026Embodied AIMESORehabilitation RoboticsExoskeletonsReinforcement LearningStroke RecoveryGait TrainingAdaptive Control

Reinforcement Learning-Adaptive Rehabilitation Exoskeletons for Post-Stroke Gait Recovery

Sim-to-Real Transfer and Human-in-the-Loop Optimization for Personalized Lower-Limb Assistance

Hass Dhia — Smart Technology Investments Research Institute

Market opportunity analysis for reinforcement learning-adaptive rehabilitation exoskeletons targeting post-stroke gait recovery. Four independent research groups have demonstrated metabolic cost reductions of 24.2% (Zhang et al., Science 2017), 23% (Slade et al., Nature 2022), 24.3% (Luo et al., Nature 2024), and 5.3-19.7% across ten activities (Molinaro et al., Nature 2024) using human-in-the-loop optimization, Bayesian optimization, and sim-to-real RL transfer on hip and ankle exoskeletons tested on human subjects. Zero commercial rehabilitation exoskeletons use RL-adaptive closed-loop control — all FDA-cleared devices (ReWalk, Ekso Bionics, Indego) operate with fixed, pre-programmed gait trajectories. Bottom-up TAM of $3.2B annually for US post-stroke gait rehabilitation, cross-checked against the $1.5B rehabilitation robotics market growing at 12.5% CAGR (Verified Market Reports, 2024). The gap between demonstrated RL-adaptive control and clinical deployment — personalization for neurological impairment profiles, manufacturing of sensor-integrated exoskeletons at rehabilitation facility volumes, and regulatory strategy for adaptive AI-controlled Class II medical devices — represents a first-mover opportunity in a device class where reimbursement pathways already exist (HCPCS K1007, CPT 97116).

2026Embodied AIMESOSurgical RoboticsAutonomous SurgeryImitation LearningLaparoscopic SurgerySoft Tissue

Autonomous Soft Tissue Surgical Systems with Learned Dexterity

Imitation Learning and Hierarchical Control for Surgeon-Independent Laparoscopic Procedures

Hass Dhia — Smart Technology Investments Research Institute

Market opportunity analysis for autonomous soft tissue surgical systems operating at Level 4 surgical autonomy. Three landmark studies — autonomous intestinal anastomosis in live porcine models (Science Translational Medicine, 2016; Science Robotics, 2022) and autonomous cholecystectomy clipping-and-cutting in ex vivo porcine gallbladders with 100% task completion across 8 unseen specimens (Science Robotics, 2025) — establish that AI-driven surgical robots can match or exceed expert surgeon consistency in soft tissue procedures. Zero commercial products exist at Level 4 autonomy; all FDA-cleared surgical robots operate at Level 1-3. ARPA-H committed $12M to the ALISS program (2024) for autonomous surgical development. Bottom-up TAM of $4.1B annually for US laparoscopic cholecystectomy alone, cross-checked against the $12.8B surgical robotics market (Mordor Intelligence, 2025). The gap between demonstrated autonomous capability and clinical deployment — regulatory strategy for AI-controlled surgical instruments, manufacturing of sensor-rich end-effectors at clinical volumes, and validated safety architectures for human-supervised autonomous operation — represents a first-mover opportunity in a device class with no commercial precedent.

2026Embodied AIMICROBioelectronic MedicineVagus Nerve StimulationNeuroimmune ModulationReinforcement LearningAutoimmune Disease

Closed-Loop Adaptive Bioelectronic Implants for Chronic Inflammatory Disease

Reinforcement Learning-Optimized Vagus Nerve Stimulation for Autoimmune Conditions

Hass Dhia — Smart Technology Investments Research Institute

Market opportunity analysis for closed-loop adaptive bioelectronic implants targeting chronic inflammatory disease. The biological mechanism — vagus nerve-mediated neuroimmune modulation — received FDA approval in July 2025 via SetPoint Medical's open-loop device for rheumatoid arthritis (Nature Medicine, 2025). Zero commercial products exist for closed-loop adaptive stimulation, where reinforcement learning algorithms optimize stimulation parameters in real time based on physiological biomarkers. Computational studies demonstrate 67% power reduction versus open-loop while preserving therapeutic efficacy (IEEE TNSRE, 2024). Bottom-up TAM of $6.2B annually for US treatment-resistant moderate-to-severe RA, cross-checked against the $25.8B bioelectronic medicine market (Grand View Research, 2023). The gap between FDA-validated biology and AI-optimized delivery — adaptive control algorithms, miniaturized closed-loop hardware, and manufacturing at clinical volumes — represents a first-mover opportunity in a device class with established regulatory precedent.

2026Embodied AIMICRODrug DeliveryMicroroboticsTargeted TherapyOncology

Magnetically Guided Microrobots for Targeted Intravascular Drug Delivery

Sub-Millimeter Autonomous Navigation for Precision Chemotherapy

Hass Dhia — Smart Technology Investments Research Institute

Market opportunity analysis for magnetically guided microrobotic drug delivery platforms. Foundational research validated in large animal models (Science 2025) demonstrates sub-millimeter capsule navigation against physiological blood flow under clinical fluoroscopy. Zero commercial products exist at the sub-millimeter intravascular scale. Bottom-up TAM of $2.5B annually for US oncology applications alone, cross-checked against the $10.7B targeted drug delivery market (Coherent Market Insights, 2025). The gap between laboratory validation and clinical deployment — AI-driven autonomous navigation, batch manufacturing, and regulatory strategy — represents a first-mover opportunity in a space where academic labs lack commercialization infrastructure.

AnestheSim paper title page
2026Reinforcement LearningAnesthesiologyDrug DosingOpen Source

AnestheSim

Gymnasium Environments for Reinforcement Learning in Automated Anesthesia Drug Dosing

Hass Dhia -- Smart Technology Investments Research Institute

Three Gymnasium-compatible RL environments for automated anesthesia drug dosing: propofol infusion control via the Marsh three-compartment pharmacokinetic model with Hill pharmacodynamic BIS prediction, remifentanil effect-site concentration targeting via the Minto model, and combined propofol-remifentanil anesthesia management using the Greco synergistic interaction surface. Includes configurable difficulty tiers with patient variability and surgical stimulation events, heuristic TCI clinical baselines, PPO RL agents, and a benchmark suite across three difficulty levels. Key finding: pharmacokinetic timescale, not task complexity, is the primary determinant of RL sample efficiency in drug dosing control. 109 tests, MIT licensed.

NeuroSim paper title page
2026Reinforcement LearningBrain-Computer InterfacesNeuroscienceOpen Source

NeuroSim

A Gymnasium Platform for Reinforcement Learning in Brain-Computer Interfaces

Hass Dhia — Smart Technology Investments Research Institute

Gymnasium-compatible RL environment suite for brain-computer interfaces with three environments modeling motor imagery decoding, intracortical cursor control, and P300 speller navigation. Includes pluggable signal models (electrode drift, fatigue, co-adaptation, noise), a conditional VAE neural surrogate, CSP+LDA classical baseline, PPO RL baseline, and a five-tier benchmark suite. 158 tests, MIT licensed.

VascularSim paper title page
2026Reinforcement LearningMedical RoboticsSimulationOpen Source

VascularSim

A Gymnasium Platform for Microrobot Navigation in Patient-Derived Vascular Networks

Hass Dhia — Smart Technology Investments Research Institute

Open-source simulation platform providing a complete stack for training RL agents to navigate blood vessel graphs: TubeTK data ingestion, three Gymnasium environments with physics-based observations (VascularNav, FlowAwareNav, MagneticNav), analytical hemodynamics and magnetic field models, a neural flow surrogate, PPO baseline agents, and a benchmark suite across 5 difficulty tiers. 139 tests, MIT licensed.

PeptideGym paper title page
2026Reinforcement LearningDrug DiscoveryPeptide EngineeringOpen Source

PeptideGym

Gymnasium-Compatible RL Environments for Therapeutic Peptide Design

Hass Dhia — Smart Technology Investments Research Institute

Three Gymnasium-compatible RL environments for therapeutic peptide design: antimicrobial peptides (AMP), cyclic peptides, and T-cell epitopes. Includes heuristic biophysical scoring models, PPO and random baseline agents, reward shaping analysis revealing mode collapse boundaries, and a benchmark suite. First systematic demonstration that per-step reward shaping magnitude determines whether RL agents learn meaningful peptide sequences or degenerate to single-residue exploitation. 125 tests, MIT licensed.

OncoSim paper title page
2026Reinforcement LearningRadiation TherapyOncologyOpen Source

OncoSim

Gymnasium Environments for Reinforcement Learning in Radiation Therapy Treatment Planning

Hass Dhia — Smart Technology Investments Research Institute

Three Gymnasium-compatible RL environments for radiation therapy treatment planning: beam angle optimization, dose fractionation scheduling, and adaptive replanning. Includes analytical pencil beam dose calculation, linear-quadratic cell survival, TCP/NTCP radiobiological models, configurable difficulty tiers, and baseline agents (random, heuristic, PPO). PPO achieves 11.7x improvement on beam selection and 15.4x on adaptive replanning over clinical heuristics. 141 tests, MIT licensed.

GlucoSim paper title page
2026Reinforcement LearningGlucose ManagementDiabetesOpen Source

GlucoSim

Gymnasium Environments for Reinforcement Learning in Glucose Management

Hass Dhia — Smart Technology Investments Research Institute

Three Gymnasium-compatible RL environments for Type 1 diabetes glucose management: basal rate optimization, meal bolus dosing, and full closed-loop insulin delivery. Includes the Bergman minimal glucose-insulin model, Dalla Man gut absorption dynamics, a CGM sensor noise model, 30 virtual patients across three age groups, configurable difficulty tiers, heuristic clinical baselines, PPO RL agents, and a five-tier benchmark suite. Key finding: composite reward functions with safety constraints are necessary to differentiate learned policies from naive baselines in glucose management RL. 117 tests, MIT licensed.

VentiSim paper title page
2026Reinforcement LearningMechanical VentilationCritical CareOpen Source

VentiSim

Gymnasium Environments for Reinforcement Learning in Mechanical Ventilation

Hass Dhia -- Smart Technology Investments Research Institute

Three Gymnasium-compatible RL environments for mechanical ventilation: tidal volume control via inspiratory pressure adjustment, PEEP optimization for oxygenation, and full ventilator parameter management for ARDS patients. Implements a single-compartment lung mechanics model coupled with a simplified gas exchange model, configurable difficulty tiers with patient variability and disease progression, heuristic clinical baselines, and PPO agents. Key finding: PPO improvement over baselines scales monotonically with action dimensionality, from 11.8% in 1D to 65.0% in 4D control. 230 tests, MIT licensed.

CardioSim paper title page
2026Reinforcement LearningCardiac ElectrophysiologyDrug DosingOpen Source

CardioSim

Gymnasium Environments for Reinforcement Learning in Cardiac Electrophysiology

Hass Dhia -- Smart Technology Investments Research Institute

Three Gymnasium-compatible RL environments for cardiac electrophysiology: pacemaker rate optimization via the FitzHugh-Nagumo model with a cardiac conduction system simulator, antiarrhythmic drug dosing using the FitzHugh-Nagumo model with single-compartment PK/PD dynamics, and defibrillation timing via the Aliev-Panfilov model with probabilistic shock success. Includes configurable difficulty tiers, heuristic clinical baselines, and PPO agents. Environments span a difficulty spectrum from learnable (drug dosing, pacing) to open-challenge (defibrillation timing). 134 tests, MIT licensed.

12

Publications

37

Domains

8

Open-Source Repos

8

Published Packages

Collaborate with us

We welcome research collaborations, dataset contributions, and open-source partnerships across any discipline. Reach out to discuss.