Resume

VLED Lab, IIT Ropar · Remote

Researching and deploying scalable AI/ML pipelines on live open-source repositories targeting critical gaps in India's education ecosystem, with direct commit authorship. Under Dr. Sudarshan Iyengar, Associate Professor, IIT Ropar.

Reaching Sky Foundation · Remote

Built a BERT-based multi-label classifier to detect mental-health disorders from user posts using attention mechanisms. Managed and optimized a responsive website with technical SEO to improve visibility and engagement.

BreakOut AI · Remote

Built an AI-powered tool for data and web insights that processes datasets, performs intelligent web searches, and uses LLMs to extract user-defined information through a streamlined interface.

Codesoft Technologies Pvt. Ltd. · Remote

Performed exploratory data analysis on large datasets to identify patterns and trends, and built machine-learning models for classification and regression problems.

Postman · Remote

Certified Student Expert with experience in API scripting, workflow validation, and test automation using Postman tools.

Developing a hybrid RNN–Transformer model to enhance language-modeling efficiency in low-resource settings — optimizing parameter use, memory footprint, and inference latency without compromising accuracy.

Fine-tuned Qwen2-VL-2B with 4-bit QLoRA (Unsloth) on IDRiD to grade diabetic retinopathy with structured clinical explanations on a single 6 GB GPU — accuracy 31.1%→66.0%, QWK 0.00→0.76. Released a 2,064-pair VQA dataset and model on Hugging Face with an ACL-format paper. Tech: PyTorch, Unsloth, QLoRA, Gradio.

End-to-end fake-news classifier (TF-IDF + linear models) reaching 0.99 F1 / 0.9995 AUC on a held-out test set — fully reproducible, unit-tested with GitHub Actions CI, Gradio demo, and a source-leakage analysis. Tech: Python, Scikit-learn, Pytest.

AI evaluation system using BERT & SBERT for semantic scoring of descriptive answers, with OCR for handwritten text — reducing manual evaluation time by 70%+. Built under Dr. Ashwini Kumar Upadhyay (REC Kannauj). Tech: Python, TensorFlow, Hugging Face, OpenCV, Flask, Streamlit.

AI-driven system to detect road hazards, lane deviations, and obstacles for biker safety using deep learning and computer vision.

BERT-based multi-label classifier detecting disorders such as anxiety and depression from text using NLP and attention mechanisms.

LLM-powered tool for dataset processing, intelligent web search, and structured information extraction through a clean interface.

YOLOv8 system for vehicle detection, speed estimation, and collision-risk prediction with real-time alerts and object tracking.

CNN-based face-recognition model using OpenCV and FaceNet embeddings for accurate identity verification.

Rajkiya Engineering College, Kannauj

CGPA 7.5. Coursework: Digital Electronics, Microprocessors & Microcontrollers, DSP, Control Systems.

IIT Madras

Mathematics for Data Science, Statistics, Advanced Python, AI & Machine Learning.

National Finalist (top 2% of 40,000+ participants) at the GUVI–HCL Buildathon for innovation and AI-driven problem-solving.