E B BENSON RIKEN BRC · 36°05′N 140°05′E · TSUKUBA, JP

FOLIO // 2026 — RIKEN LOTUS-JST RESEARCH FELLOW

E B Benson

Investigating the computational grammar of AI Virtual Cells, single-cell gene regulation, and autonomous agentic reasoning.

At RIKEN BioResource Research Center in Tsukuba, I work with Naoki Kubo’s group on foundation models for cellular state transitions and single-cell omics, while completing an Integrated M.Sc. in AI & Data Science at CUSAT. Previously engineered agentic document processors at SpectraAI (Singapore) and researched stochastic time-series models at IIT Bombay.

§ I · CELLULAR COMPUTATION

Single-cell multi-omics, gene regulatory network inference, and predictive AI Virtual Cells for iPSC regenerative medicine.

§ II · AGENTIC SYSTEMS

Multi-agent communication protocols (ACP), type-safe agent frameworks (PydanticAI), LoRA fine-tuning, and document RAG.

§ III · VISION & DYNAMICS

Unified real-world image restoration across multi-degradations (ECCV LoViF) and econometric vs. attention sequence models.

GENOMICS COMMUNICATIONS JOURNAL 2026

Artificial Intelligence and Single-Cell Omics: From Gene Regulation to Virtual Cells with Applications in iPSC-Based Regenerative Medicine

Synthesizes how multi-scale foundation models decode single-cell transcriptomic and epigenomic regulation to simulate cellular trajectories in silico, addressing reproducibility and lineage control in induced pluripotent stem cell (iPSC) therapies.

AI Virtual Cells Single-Cell Omics Gene Regulatory Networks iPSC Regenerative Medicine
SPRINGER NATURE METEOR JOURNAL · ECCV 2026 WORKSHOPS 2026

The Second LoViF 2026 Challenge on Real-World All-in-One Image Restoration: Methods and Results

JUL 2025 — NOV 2025 SpectraAI · Singapore

AI Intern

Architected and built an Agentic Document Processor for autonomous multi-stage document parsing, schema extraction, and verification.

W—01

Parkinson’s Disease Prediction from Blood Gene Expression (GEO)

REPOSITORY ↗

End-to-end bioinformatics pipeline pairing deep autoencoders with supervised classifiers on Gene Expression Omnibus (GEO) cohorts. Applied normalization, Boruta/RFE feature selection, and dimensionality reduction to reach 0.80 AUC on early-stage Parkinson’s detection.

Python / PyTorch / Scikit-learn / GEOquery

W—02

Cognitive LLM for Algebraic Question Generation

REPOSITORY ↗

Parameter-Efficient Fine-Tuning (LoRA) of large language models over structured algebra corpora, paired with a custom dataset synthesis mechanism for generating valid LaTeX-formatted mathematical problems.

Hugging Face Transformers / LoRA / LangChain / PyTorch

W—03

Automated Customs Declaration Intelligence

LOCAL RAG

Structured table and document layout extraction from complex PDFs via Docling, dense vector indexing in PostgreSQL, and local LLM inference through LM Studio for automated customs form synthesis.

Docling / PostgreSQL / LM Studio / Vector Retrieval

W—04

Forecasting Financial Variations via Autoregressive & Attention Models

ECONOMETRICS

Comparative time-series study of cryptocurrency volatility and INR–USD currency fluctuations, benchmarking parametric formulations (AR, ARIMA, TAR) against modern attention-based sequence architectures.

R / forecast / tseries / Attention Architectures

W—05

Automated Parking Vendor Allocation System

FIELD TRIALS

Co-created a real-time, vendor-assistive smart mobility allocation system deployed in live trial phases in India to optimize urban traffic flow and vendor load balancing.

React Native / Node.js / Firebase / Google Cloud

FELLOWSHIPS & DISTINCTIONS

2027
LOTUS ASPIRE JST Fellowship Recipient · Japan Science and Technology Agency
2026
LoViF 2026 Image Restoration Challenge Top 20 out of 1,740 participants · ECCV 2026
2026
LOTUS-JST Fellowship Recipient · RIKEN & JST
2026
GATE Data Science & AI (DA) Qualified · National Graduate Aptitude Test in Engineering
2024
FOSSEE Summer Fellowship Recipient · Indian Institute of Technology Bombay