01Accepted paper · ASONAM 2026
LLM Stance Perception
Women’s Safety Narratives
Measure how stance varies across large-scale social-media discussion of women’s-safety cases in India.
Qwen3.5-9BLoRAMann–Whitney UChi-square
Open case study 01 Data scientist · computational researcher
Charishma
Reddy Yerra.
I build statistical and machine-learning systems for messy, consequential data—from a 4.3M-comment stance-analysis pipeline to Bayesian models for mapping disease variants to candidate genes.
4.3M+
raw comments processed
351,501
curated observations
0.7450
macro-F1
16
cases analyzed
∂L/∂θ → 0, but the questions keep moving.
02 f(me)
Evidence in layers. Uncertainty left visible.
My work sits where data pipelines, statistical tests, machine learning, and scientific questions meet. I care about the part before the model—what gets measured, cleaned, compared, and discarded—and the part after it: evaluation, error analysis, and what the result can actually support.
More about my workpipelines
X_raw → X_curatedmodels
argminθ L(θ)research
P(H | evidence)03 argmax(project impact)
Selected work
Three research systems that show how I frame problems, combine evidence, and design evaluations.
02Nearing completion
BayesV2G
Bayesian Variant-to-Gene Prioritization
Prioritize candidate causal genes for cardiometabolic traits from disease-associated variants.
Bayesian inferenceGWAScis-eQTLcis-pQTL
Open case study 03Nearing completion
CoreGene-Bayes
Probabilistic Core-Gene Discovery
Identify candidate core disease genes by modeling convergence of genetic perturbations across molecular pathways.
Bayesian modelingMendelian randomizationMolecular QTLGene networks
Open case study 04 Σ experience
From GPU kernels to language models.
Aug 2025 – Present
Graduate Research Capstone
Rochester Institute of Technology
Large-scale discourse and stance analysis across women’s-safety cases in India.
P(stance | narrative)Aug 2025 – Present
GCCIS Technical Assistant
Rochester Institute of Technology
Academic computing support, software implementation, and resource analysis.
uptime → 1Jan 2025 – Present
Research Assistant
RIT School of Interactive Games and Media
Quantitative analysis and research communication for instructional-media work.
H₀ ↔ H₁Jan 2024 – May 2024
Innovation Intern
Forge Innovation and Ventures
User-engagement analysis, KPI reporting, and product-strategy research.
engagement ↑ 25%Jun 2023 – Aug 2023
High-Performance Computing Intern
National Atmospheric Research Laboratory
GPU-accelerated matrix operations and kernel benchmarking.
O(n³) // GPU05 H₀ vs H₁
Accepted · ASONAM 2026
A short paper built from 4.3 million social-media comments.
The work combines large-scale data curation, nonparametric and categorical tests, LoRA adaptation, held-out evaluation, and high-confidence error auditing.
Research recordaccuracy.7447
macro-F1.7450
MCC.6634
n held out3,000
06 X ∈ ℝⁿ
A working feature space
No percentage bars. The tools below are tied to projects, research, or professional work in the attached résumés.
L
Programming
Python · R · SQL · Bash · C
M
Machine learning & NLP
Scikit-learn · XGBoost · PyTorch · LLMs · LoRA / fine-tuning
θ
Bayesian & statistics
Bayesian inference · Hierarchical models · Posterior estimation · Hypothesis testing · Regression
G
Statistical genetics
GWAS · eQTL / pQTL analysis · Colocalization · Variant-to-gene mapping · Mendelian randomization
X
Computational genomics
Molecular QTL integration · Transcriptomics · Spatial gene expression · Gene-set enrichment · GTEx
H
Scientific & HPC
PyMC · ArviZ · SciPy · NumPy · Pandas
07 prior knowledge
Education
Aug 2024 – May 2026
M.S. Data Science
Rochester Institute of Technology
Rochester, NYGPA 3.710 / 4.0
Sep 2020 – May 2024
B.E. Computer Science and Engineering
RMK Engineering College
Chennai, IndiaGPA 3.825 / 4.0
P(reply | context) ↑