Skip to content
View memnaaaaa's full-sized avatar

Block or report memnaaaaa

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
memnaaaaa/README.md

hi, I'm Mohini Tiwari

MS CS @ Georgia State · software engineer · ML researcher
I build across the stack: from backend systems to on-device AI inference.
previously data analyst @ PreProd Corp

🍳 home cook · 🎸 guitarist · ✈️ solo traveller · 🏋️ powerlifter · 🍵 matcha


what I work with

Languages   Python · Java · C · TypeScript · Bash
ML/DL       PyTorch · TensorFlow · HuggingFace · Transformers · scikit-learn
Data        pandas · NumPy · PostgreSQL · Redis · MariaDB · KNIME
Backend     FastAPI · REST APIs
Infra       Linux · Docker · Git · GitHub Actions · MLflow
Practices   OOP · CI/CD · MLOps · system design
Workflow    GitHub · Jira

right now

  • MS CS @ Georgia State — GPA 4.15 / 4.30
  • research: distributed AI · efficient inference · computer vision
  • open to summer 2026 internships — SWE, ML, or data
  • 📍 Atlanta, GA

featured projects

⚡ PADQ — Prompt-Aware Dynamic Quantisation Zero-shot routing system that dynamically switches between INT4 and FP32 LLM inference based on prompt complexity — optimised for edge deployment.
Achieved 72.6% energy reduction, 39.1% latency improvement, and 1.5–2.2× throughput gains on mobile hardware.
TinyLlama · Qwen3-0.6B · iPhone A16/A18 · on-device inference

🔬 RobustCAM — Faithful Grad-CAM for Lung CT Classification ResNet50 fine-tuned on IQ-OTH/NCCD (normal / benign / malignant), achieving val_acc=0.8813 vs. published baseline of 0.85. Fuses Grad-CAM across augmented views and cross-validates with LIME and SHAP using a 9-metric faithfulness suite.
ResNet50 · Grad-CAM · LIME · SHAP · MLflow · IQ-OTH/NCCD


background

I came into CS through data — spent time as a data analyst before moving into research and engineering. That path shaped how I think: I care about whether systems are measurable, reproducible, and actually useful outside a notebook. I like problems that sit at the boundary between research and deployment.


"building things that make sense on the edge"

email · linkedin

Pinned Loading

  1. fed-personalized-conformal fed-personalized-conformal Public

    Personalized federated conformal prediction using adaptive per-client global-local threshold weighting to improve coverage efficiency under non-IID data.

    Python

  2. robustcam-xai robustcam-xai Public

    Augmentation-fused Grad-CAM with LIME/SHAP cross-validation and a 9-metric faithfulness suite for explainable lung CT classification using fine-tuned ResNet50 on IQ-OTH/NCCD

    Python

  3. target-rec target-rec Public

    ML-powered product recommendation system that analyzes guest reviews using sentiment analysis (DistilBERT) to surface bestselling products by category and generate actionable insights for underperf…

    Jupyter Notebook

  4. ItsHariii/UGA11 ItsHariii/UGA11 Public

    TypeScript

  5. CSC8830-ComputerVision CSC8830-ComputerVision Public

    Python

  6. CSC8980-DistributedAI CSC8980-DistributedAI Public

    Python