I'm an AI/ML engineer and researcher interested in building reliable, scalable machine learning systems.
My work spans deep learning, uncertainty estimation, conformal prediction, geospatial ML, computer vision, and intelligent agents. I enjoy taking ideas from research papers and turning them into working systems.
Currently exploring:
Machine Learning Deep Learning Computer Vision Geospatial AI Conformal Prediction LLM Agents
Accepted at IEEE ICDM 2026 β Main Research Track
A research project on uncertainty-aware forecasting over supply-chain hypergraphs, combining hypergraph structure with conformal prediction to produce reliable prediction intervals.
Key ideas:
- Hyperedge-aware representation of supply-chain relationships
- Quantile-based forecasting
- Adaptive online calibration
- Uncertainty-aware prediction intervals
- Evaluation across different structural attributes
Paper: Hyperedge-Aware Conformal Prediction for Supply-Chain Forecasting
π Repository
Python PyTorch Hypergraphs Conformal Prediction Time Series Forecasting
Published at MIGARS 2026 β IEEE GRSS
Developed a satellite-driven ML pipeline for predicting surface soil moisture at 100m resolution using remote sensing and geospatial covariates.
The system combines:
- Sentinel-1 radar imagery
- Google Earth Engine
- Ensemble learning
- PyTorch MLP
- XGBoost / Random Forest
- Explainability with SHAP
π Repository
Python PyTorch Google Earth Engine XGBoost SHAP
| Achievement | Result |
|---|---|
| π₯ IEEE ICDM 2026 | Accepted β Main Research Track |
| π₯ TenzorX National AI Hackathon | Top 150 of 7,000+ teams |
| π₯ Hero Campus Challenge | Top 5 of 200,000+ participants |
| π MIGARS 2026 | Published research |
π CareerSight
Research-oriented career outcome prediction system combining causal inference, contextual bandits, survival analysis, and federated learning.
π Top 150 / 7,000+ teams at the TenzorX National AI Hackathon.
Python Causal Inference Reinforcement Learning Federated Learning
π Car Damage Detection
Multi-stage vehicle inspection pipeline using instance/semantic segmentation and vision-language models for damage detection, severity classification, and vehicle-component identification.
π Top 5 / 200,000+ participants β Hero Campus Challenge.
Python Computer Vision Mask2Former VLM
π° Finance Agent
An AI-powered finance agent focused on combining LLM reasoning with financial data and tool-based workflows.
Python LLM Agents RAG APIs
Implemented the core Transformer architecture from first principles to understand:
- Self-attention
- Multi-head attention
- Positional encoding
- Encoder / decoder architecture
- Masking
- Training pipeline
Python PyTorch Transformers
| Project | Description |
|---|---|
| π¬ RAG Doc/Video Chat | Retrieval-augmented systems for documents and video transcripts |
| π¬ Movie Recommendation System | Content-based and similarity-driven recommendation |
| π€ AI Mock Interview Tool | LLM-powered interview simulation and feedback |
| π‘ Signal Modulation Classifier | Deep learning for radio-signal classification |
| π€ Student Query Chatbot | REST-based conversational AI application |
π§ vishalpandey1623@gmail.com
Open to ML research, research internships, applied ML opportunities, and collaborations.
β Check out the repositories above if you're interested in the work.