class ShreyanshGupta:
role = ["SDE / SWE", "AI-ML Engineer", "Applied Researcher"]
location = "NIT Patna, India"
focus = ["Deep Learning", "Computer Vision", "NLP", "Graph ML"]
engineering = ["Full-stack web", "MLOps / DevOps fundamentals"]
mindset = "Research the idea → prototype it → ship it to production"- 🔬 I build explainable, lightweight deep-learning systems — from terrain-recognition CNNs to hybrid graph + language embeddings for citation sentiment analysis.
- ⚙️ I work across software engineering and ML — data → training → API → deployed app, and the full stack around it.
- 🌐 Comfortable across the MERN / TypeScript stack and the Python ML ecosystem (PyTorch, scikit-learn).
- 🚀 Currently going deeper on applied deep learning, model deployment, and reproducible ML pipelines.
- 🏆 Amazon ML Summer School 2025 — selected among the top 3,000 of 60,000+ applicants.
- 🤝 Open to SDE / SWE and AI-ML roles — 2026 internships and 2027 new grad.
| Venue | Paper |
|---|---|
| IEEE MIGARS 2026 | Crop Phenology Classification: A Multi-Paradigm Benchmark from Traditional ML to Hybrid Mamba–Transformer Networks |
| MIND 2025 | Hybrid Graph–Language Embedding Framework for Citation-Aware Sentiment Analysis |
|
High-accuracy, explainable, lightweight CNN for terrain classification — designed to be deployable on constrained hardware without trading away interpretability.
|
MIND 2025 paper. Fuses Node2Vec citation-graph structure with BERT sentence embeddings through a GNN — 90.56% accuracy on the ACL Citation Sentiment Corpus, beating text-only baselines.
|
|
Full-stack MERN streaming app — TMDB API, JWT auth, search history & trailers, with a live Netlify frontend.
|
Civic-infrastructure PWA with a LangGraph agent crawling OpenStreetMap and local news, zero-token semantic search over ChromaDB, and 2SFCA accessibility heatmaps. India Innovates 2026 finalist.
|
|
IEEE MIGARS 2026 paper, sponsored by RAC-S, ISRO. Winter wheat growth-stage classification from Sentinel-1 SAR and Sentinel-2 optical time series over 204 fields in Bihta, Bihar — benchmarking eight models across three modality configurations, reaching macro-F1 0.7765.
|
|
👾 Pac-Man munches through my commits — regenerates every 24h via GitHub Actions
→ shreyansh-gupta-ml.netlify.app
"The best way to predict the future is to build it — then measure it, and iterate."