A selection of projects at the intersection of AI/data science and real-world impact.
For full publications, see my Google Scholar. Connect on LinkedIn.

Contextual Embeddings for Emotion & Meaning (BERTNN / ACT)

Role: Lead developer (PhD research)
What it does: Integrates transformer models with Affect Control Theory to expand affective lexicons and track emotion dynamics in messaging.
Methods: BERT, lexicon induction, sequence modeling
Outputs: Journal article in Sociological Methodology (2024) and related conference papers
Funding/acknowledgment: Includes ARO-supported work (W911NF1710509)
Links: Journal article · UMAP’21 paper · IEEE SysCon’21 (emoji/embeddings)

Mindful Driving: Safe & Sustainable Fleet Analytics

Role: PI / Technical lead
What it does: Uses telematics + ML to identify risky/inefficient driving and delivers targeted eco-driving training.
Impact: Recognized as NAFA “Best Special Project in North America” (2021)
Methods: Telemetry pipelines, feature engineering, behavioral interventions

Collision Prediction & Prevention in Contact Sports (RFID + Haptics)

Role: Lead researcher
What it does: Predicts imminent player-to-player collisions via RFID tracking and alerts athletes with haptic feedback.
Methods: RFID localization, real-time thresholds, prototype evaluation
Outputs: Refereed proceedings chapter
Links: Conference chapter

Sensor-free Microclimate Monitoring Using Existing LoRaWAN Signal Characteristics

Authors: F. Nikseresht, V. A. L. Sobral, M. Mostafavi, B. Campbell
What it does: Estimates local microclimate conditions without dedicated sensors by exploiting signal characteristics from existing LoRaWAN infrastructure.
Methods: Wireless signal feature extraction, environmental inference, low-power IoT
Venue: International Workshop on Environmental Sensing Systems
Links: Paper

A Tale of Two Metrics: Finance vs. Polling in U.S. Primaries

Role: Lead analyst
What it does: Uses joinpoint regression to study the relationship between campaign contributions and polling performance (2020 Democratic primaries).
Methods: Time-series modeling, joinpoint analysis, data integration
Links: IEEE SysCon’21 paper

Expectation Modeling in Macroeconomic Systems

Role: Lead author
What it does: Proposes a predictive multi-agent approach to incorporate rational expectations while addressing stability and uniqueness limitations.
Methods: Agent-based modeling, control/optimization, macro dynamics
Funding: Iranian National Science Foundation (INSF)

Sleep Technology Feasibility in Older Adults with MCI

Role: Co-author / data science support
What it does: Assesses feasibility of technology use in an internet-delivered sleep intervention for older adults with mild cognitive impairment.
Links: Alzheimer’s & Dementia (conference abstract)

Mortgage-Backed Securities (MBS) Analytics at VRS

Role: Investment Officer (applied analytics)
What it does: Builds and validates models that integrate borrower/loan attributes with market indicators to support MBS risk/return decisions.
Methods: ML for tabular data, credit/survival modeling, scenario analysis
Note: Professional work; summaries and public-safe visuals available on request.