ML & applied-mathematics researcher working at the intersection of spectral methods, regularization theory, graph signal processing, and learning-to-rank — with an applied bent toward demand forecasting and information retrieval.
I build principled, math-forward methods and turn them into working systems. Recurring threads in my work:
- Spectral & graph signal processing — Laplacian/eigenstructure methods, signals and flows on (directed) graphs
- Regularization theory — well-posed, parameter-insensitive formulations for recovery and ranking
- Learning-to-rank & information retrieval — multi-metric rank prediction, search, and retrieval
- Applied forecasting — demand and time-series modeling grounded in the above
Day to day I work in my own Python and Scala research stacks.
- Latent Aspect Detection from Online Unsolicited Customer Reviews — an unsupervised method for extracting latent aspects from reviews. Forouhesh, Mansouri, Fani (2022). arXiv:2204.06964 · code
- Tracking Legislators' Expressed Policy Agendas in Real Time — real-time computational-social-science pipeline for policy-agenda tracking. code
| 🎁 Project | 📝 What it does | ⭐ Stars | 🍴 Forks |
| Latent Aspect Detection | Unsupervised extraction of latent aspects from customer reviews (arXiv:2204.06964) |  |
 |
| Tracking Policy Agendas | Real-time tracking of legislators' expressed policy agendas |  |
 |
| Persian POS Tagger | Conditional Random Field framework for Persian part-of-speech tagging |  |
 |
| Twitter Sentiment Engine | BERT-based Persian Twitter sentiment-analysis framework |  |
 |
| Persian Relatio | Persian reimplementation of Elliott Ash's Relatio narrative-extraction framework |  |
 |
| Inverted Index | A simple search engine built from scratch |  |
 |
Languages
ML & NLP
-1BA0E2?style=flat-square&logo=apachespark&logoColor=white){kind=icon}
Distributed & Streaming
Infrastructure & Ops
