CSE undergrad @ MSIT · Machine Learning Engineer · RAG, Agentic AI & Deep Learning
from dataclasses import dataclass, field
@dataclass
class Jiya:
role: str = "ML Engineer in the making"
focus: list = field(default_factory=lambda: [
"natural language processing","retrieval-augmented generation",
"AI safety & red teaming", "applied deep learning", "DSA"
])
building: str = "hybrid-retrieval RAG systems that actually cite the right page"
learning: str = "agentic workflows, eval harnesses, inference optimization"
motto: str = "turn the theory into something that ships"- ML Summer Intern at the Airports Authority of India — neural nets from scratch, plus retrieval and latency work on an enterprise RAG system.
- AI/ML Intern at DomaiynLabs — adversarial red teaming research and a prompt-injection firewall served over FastAPI.
- LinkedIn CoachIn mentee — selected among the top 100 from 21,000+ applicants, and won the internal hackathon.
- 9.19 / 10 CGPA, B.Tech CSE @ Maharaja Surajmal Institute of Technology, graduating 2028.
Languages
AI / ML
Data & Vector Stores
Backend & Tools
| Project | What it is | Stack | Highlight |
|---|---|---|---|
| Aviation Safety RAG Assistant | Hybrid-retrieval assistant over AAI's ATS manual and ATMC circulars | Python · LangChain · Llama 3.1 · ChromaDB · Cross-Encoder · Streamlit | BM25 + vector + cross-encoder reranking across 670+ pages, with PyMuPDF/RapidOCR ingestion for scanned legacy PDFs and a Pytest regression suite guarding retrieval accuracy |
| SpeakEZ | Real-time AI coaching platform simulating workplace conversations | FastAPI · Node.js · Gemini Live API · PyTorch · sentence-transformers · MediaPipe · Prisma | Low-latency voice loop plus a multi-modal scoring pipeline — embeddings, speech fluency, acoustic confidence and live MediaPipe nudges — feeding personalized insights |
| Marketplace Product Image Classifier | Large-scale product image categorization | Python · TensorFlow · MobileNetV2 | 110,000+ images across 42 classes (10GB), transfer learning to 72.4% accuracy with imbalance handling that added 5% validation accuracy |
Machine Learning Intern — Airports Authority of India · Jun 2026 – Aug 2026 Benchmarked a custom CNN against transfer learning for image classification, cutting convergence time 5×, and improved an enterprise RAG system with BM25 retrieval, reranking, digital-text detection to skip needless OCR, and lazy-loaded model caching.
AI/ML Intern — DomaiynLabs LLP · Apr 2026 – Jul 2026 Researched adversarial testing for Bayora, an automated red-teaming platform, and shipped a firewall layer: a binary classifier for adversarial prompt detection served through a FastAPI inference backend for real-time LLM protection.
If you've scrolled this far, you might as well say hi.