Hi, I'm Himanshu Gupta
💻 ☕ ❤️
20 | Data Science | Software Engineering |
Artificial Intelligence
About
My Introduction
I’m Himanshu Gupta, a Full-Stack and AI Developer from India with 2 years of experience building scalable products for startups and contributing to open source. I’ve led full-stack development, published ML research with NAOJ (Japan), and built 20+ MVPs including RAG systems, LLM fine-tuning pipelines, and real-time AI applications using LangChain, TensorFlow, Node.js, and FastAPI. I work with Big Data technologies ( Hadoop, Kafka, Spark) and have strong foundations in system design, distributed systems, and high-performance architectures.
Work Experience
My work experiences across different companies and roles.
PropertyLens.in
WorkingJune 2025 - August 2025
Pune, India (Onsite)
Full Stack Lead Engineer
Technologies & Tools
- Developed and optimized the frontend using React.js and Next.js. Created and maintained the backend and database using Node.js and PostgreSQL. Transitioned to a React Native codebase, enabling shared development across web and mobile platforms (Project: Sidhali), focusing on component reusability.
- Addressed architectural constraints by maintaining separate components for web and mobile within the same repository, ensuring responsiveness and operational continuity. Proposed and implemented a new modular approach separating web and mobile codebases while utilizing a shared backend and database.
- Developed secure login, signup, and 2FA authentication systems integrated with thecompany’s database. Contributed to the PropertyLens School project and integrated it with the homepage. Deployed backend and database systems on Google Cloud and frontend on Vercel, ensuring full-stack deployment readiness.
- Maintained version control through structured GitHub repositories with CI/CD pipelines. Prioritized system responsiveness and architectural scalability in all deliverables.
NAOJ (National Astronomical Observatory of Japan)
December 2024 - April 2025
Machine Learning Research Intern
Featured Activity
Skills
My technical & other skills
Projects
My minimum viable products & prototypes
Portfolio Optimisation
Portfolio optimization of 7 tech stocks and S&P 500 using CAPM, Monte Carlo simulations, and SciPy optimization. Achieved max Sharpe ratio of 1.48 with Efficient Frontier visualization.
Fake News Classification
NLP text classification using LSTM and Conv1D-BiLSTM models with word embeddings, and traditional ML models with TF-IDF and Count Vectorizers. Achieved up to 93.18% accuracy.
Stock-Price-Prediction-of-Tata-Global-Beverages
Stock price prediction using Moving Average, KNN, ARIMA, and LSTM (RMSE: 24.43). Feature engineering with EMA & MACD achieved 2.49 RMSE using Linear Regression.
Research
My research publications, conference contributions, & patents
Gamma-Ray Burst Light Curve Reconstruction: A Comparative Machine and Deep Learning Analysis
Conference: Presented at Journal of High Energy Astrophysic
Link: https://arxiv.org/abs/2412.20091
Author(s): A. Manchanda, A. Kaushal, M. G. Dainotti, A. Deepu, S. Naqi, J. Felix, N. Indoriya, S. P. Magesh, Himanshu Gupta, K. Gupta, A. Madhan, D. H. Hartmann, A. Pollo, M. Bogdan, J. X. Prochaska, N. Fraija, D. Debnath
This research enhances the use of Gamma-Ray Bursts (GRBs) as cosmological tools by reducing dispersion in their light curve parameters, especially in the plateau phase critical for the Dainotti relation. Using nine ML models to reconstruct GRB light curves, MLP and Bi-Mamba significantly outperformed others in reducing uncertainty and improving prediction accuracy. These improvements strengthen GRBs' reliability as standard candles and support redshift estimation through machine learningg.
Review of Methods for Weapon Detection Using X-ray Imaging
Conference: Presented at International Journal of Advanced Computer Science and Applications
Link: https://drive.google.com/file/d/1-KJ8eLcah9juS4L9pXSgU2CnYhWvfOdd/view
Author(s): Sita Yadav, Anshu Gupta, Neha Gupta, Rahul Sawant, Sai Kalyan, Himanshu Gupta.
1. Analyzed Visual Transformers, LSTM networks, and advanced models like YOLO-V5 and F-RCNN with attention mechanisms, addressing occlusion, class imbalance, and dataset challenges with LPIXray and synthetic data. 2. Investigated few-shot and zero-shot learning, dual-energy X-ray, and spectral imaging, proposing new approaches to advance material identification and screening techniques.
