MS Computer Science @ University of Dayton — Graduating Dec 2026
- ALK+ Trajectory Dataset (master's thesis — in progress) — Building a structured, validated clinical-trajectory dataset for ALK-positive lung cancer, paired with an interpretable model that forecasts patient trajectory. Focused on reproducibility and honest, non-circular evaluation. (Full write-up and results to follow on completion.)
- Aegis (in progress) — A disease-agnostic safety-evaluation harness that measures whether clinical-triage LLMs safely handle dangerous cases — dangerous-under-triage and context-omission metrics, driven by YAML-configurable disease rules.
Languages
ML / AI
Backend
Cloud / DevOps
Data / Viz
| Project | Stack | What it does |
|---|---|---|
| 🎯 OneShot · live ↗ | Python · LLMs · Gemini · Claude · OpenAI | Local-first job pipeline — scrapes listings, LLM fit-scores each, writes ATS-optimised resumes + cover letters per job via a score→rewrite loop. +16 pts avg ATS lift, 88% fit accuracy |
| 🧠 Sage · live ↗ | FastAPI · Claude · PostgreSQL · Docker | AI-first K–10 school ERP — Claude proposes actions as reviewable cards, nothing runs without approval. 21+ domain agents, nightly insights, fee-risk scoring, full audit trail |
| 🍱 VendorVault · live ↗ | Python · Flask · SQLite · WhatsApp Webhooks | Live cloud-kitchen ops platform — WhatsApp order parsing, cash-in-hand tracking, weekly float cycle, mobile-first over WiFi |
| 🧬 ALK+ Trajectory Dataset (thesis — in progress) | Python · Machine Learning · Data Curation | Structured, validated clinical-trajectory dataset for ALK+ lung cancer + an interpretable model forecasting patient trajectory. Reproducible, honestly evaluated |
| 🛡️ Aegis (in progress) | Python · Anthropic Claude · FastAPI · pytest | Disease-agnostic safety-evaluation harness for clinical-triage LLMs — measures dangerous under-triage and context-omission via YAML-configurable disease rules |
| 📊 Rippl-Predict | Python · XGBoost · SHAP · Plotly · FastAPI · Docker | Article virality forecasting — dual regression/classification, 3D PCA dashboard, SHAP explainability, Tableau reporting (ROC-AUC 0.68–0.73) |
| 👁️ RetinaGuard | PyTorch · EfficientNet-B4 · Flask · Grad-CAM | Clinical DR screening — 5-level grading, CLAHE preprocessing, Grad-CAM, dual portals (0.82 acc / 0.89 QWK). Built on my ISSN 2321-2152 publication |
| 🔍 Auto-Lens | YOLOv12 · EasyOCR · ONNX · FastAPI · React Native | License-plate recognition + NHTSA VIN/recall lookup — DPPA-safe, runs on mobile and web |
| 🛒 SmartCart | Java · Android · ML Kit · CameraX | Grocery app with ML Kit barcode scanning, voice add via shake gesture, per-list budgets, Material 3 |
| 🤖 CRM Automation Suite | Python · pandas · rapidfuzz · Three.js · Chart.js | Outlook × PracticePanther reconciliation — 6-tier fuzzy matching with a routing reason per record, styled Excel output + a 3D analytics dashboard |
Graduate Researcher — University of Dayton (Aug 2025 – Present) My master's thesis: building a structured, validated clinical-trajectory dataset for ALK-positive lung cancer and an interpretable model that forecasts patient trajectory — with a focus on reproducibility and honest, non-circular evaluation.
Data Scientist Intern — CSIR-IICT (Jan – Aug 2023) Built a CNN + LSTM hybrid model for mosquito density forecasting to support vector-borne disease surveillance. This is where I learned to work with real biomedical data under research constraints.
IT Consultant Intern — GVS IT Solutions (Jul 2023 – May 2024) Worked across client-facing tech deployments and internal tooling — which is what pushed me toward building systems that actually get used rather than just demoed.
Data Science Apprentice — Ethical Edufabrica (Sep 2022) Early hands-on exposure to applied ML workflows and data pipelines.
Diabetic Retinopathy Detection using Deep Learning ISSN 2321-2152 · Co-authored with Niharika Raghunandan A CNN-based pipeline for automated DR detection and grading. This paper is the direct foundation for RetinaGuard — I took the research and turned it into a deployable clinical screening platform with Grad-CAM explainability, dual portals, and a GPU-free Simulation Mode.
In progress: ALK+ Trajectory Dataset — master's thesis (structured, validated clinical-trajectory dataset + interpretable trajectory model). Write-up to follow.
- President, Indian Student Association @ University of Dayton (2025 – Present)
- Director General, VBIT Model United Nations (2022 – 2023)
- Lead Organizer, TEDx VBIT (2022)
📧 tgauravsingh07@gmail.com · 📍 Dayton, OH · 🔗 Portfolio · LinkedIn
📌 Open to full-time roles in Applied AI, ML Engineering, Data Science, and Full-Stack AI Development