B.Tech (Hons.) CSE with Major in AI & ML @ MUJ
Researcher @ IAIRO | Apprentice @ IISERK | Ex-ML Researcher @ IIT-H Vigil Labs | Springer Nature Published Author | 3× Dean's List
-blue?style=flat-square){kind=icon}
Who I am:
- 🔬 Researcher @ IAIRO (Neurosymbolic AI), Apprentice @ IISERK (LLM Steering and Interpretability), and Ex-ML Researcher @ IIT-H Vigil Labs (Federated Learning)
- 📄 Author — Published by Springer Nature and SN Computer Science [Q1 - Top 25% in field]
- 🏆 3× Dean's List — Excellence in Academics & Off-campus Achievements
- 🎓 Pursuing B.Tech (Hons.) CSE with Major in AI & ML @ Manipal University Jaipur
What I build:
- 🧠 Building at the intersection of ML, NLP, Computer Vision, and Generative AI
- 🔍 Working with Neurosymbolic AI, KGs, LLM Steering RAG pipelines, Federated Learning, LLM fine-tuning
- 📈 Exploring SLMs, VLMs, model optimization, and LLM deployment at scale
Let's connect:
- 🤝 Open to collaborations on AI research, impactful projects, and roles in Applied ML.
|
Research published at ICDEC 2024 (Springer Nature). Engineered a robust ML system for barbell exercise classification and repetition counting using MetaMotion sensor data, achieving over 90% accuracy through comprehensive feature engineering and outlier detection pipelines for precise human activity recognition. Keywords: Computer Vision, Human Activity Recognition, Sensor Data, Feature Engineering, Classification |
|
Production-ready local RAG Text-to-SQL chatbot that converts natural language into safe PostgreSQL queries with 84% SQL accuracy (80% end-to-end) and ~2.5s latency. Features schema-aware retrieval (ChromaDB), code-specialized LLMs via Ollama (Qwen2.5-Coder / CodeLlama), SQL validation & security hardening, and 64 automated tests with RAGAS evaluation. Stack: Python, PostgreSQL, Ollama, Chainlit, ChromaDB, nomic-embed, pytest |
|
High-performance FastAPI microservice achieving 90% reduction in prediction latency (50ms → 5.4ms). Improved False Positive precision from 6% to 78% using class-weighted XGBoost evaluated via MLflow. Full production observability via Prometheus + Grafana, Dockerized for AWS deployment. Stack: Python, FastAPI, XGBoost, MLflow, Docker, AWS, Prometheus, Grafana |
RAG pipeline with sub-second query latency and 95% relevance accuracy across 12,000+ anime entries using LangChain, Groq LLM, and ChromaDB. Automated ETL with HuggingFace Sentence Transformers. Deployed on GCP with Kubernetes, maintaining 99.5% uptime with Grafana monitoring. Stack: Python, LangChain, Groq LLM, ChromaDB, Streamlit, Docker, Kubernetes, GCP |
🔬 Research Experience
🏛️ AI Research Intern — Indian AI Research Organization (IAIRO) (Mar 2026 – Present)
- Conducting cutting-edge research in Neurosymbolic AI and knowledge integration
- Developing frameworks that combine neural networks with symbolic reasoning for enhanced interpretability
- Contributing to foundational research in AI reasoning and knowledge representation
🏛️ AI Research Associate — IISER KOLKATA (Apprenticeship) (Mar 2026 – Present)
- Engaged in advanced research on LLM Steering and Interpretability
- Exploring techniques to guide and control large language model behavior
- Advancing understanding of LLM decision-making mechanisms and transparency
🏛️ ML Research & Development Intern — VIGIL Labs, IIT Hyderabad (Apr 2025 – Jul 2025)
- Engineered a decentralized Federated Learning model for medical image classification & segmentation
- Surpassed baseline test accuracy by 20% on complex non-IID real-world medical data
- Reduced global communication round time by 45% and utilized 55% fewer resources than baselines
- Orchestrated secure ML workflows to address data heterogeneity and strict distributed data privacy requirements
🏆 Honors & Awards
- 📄 2× Springer Nature Published Author — SN Computer Science [Q1 - Top 25% in field] & ICDEC 2024
- 🎓 Dean's List for Excellence in Academics (Highest GPA) — Manipal University Jaipur
- 🏅 2× Dean's List for Excellence in Off-campus Achievements — Manipal University Jaipur
- 🤝 Strong academic foundation in CS & AI combined with hands-on industry and research impact
☁️ MLOps & Deployment Expertise
End-to-end experience taking AI projects from development to large-scale production:
- MLOps: MLflow, DVC, DAGsHub, Airflow, Astro Airflow, TaskFlow, BentoML, Grafana
- LLMOps: LangSmith, LangServe, LangGraph, LangChain
- Cloud: AWS (S3, EC2, IAM, RDS), GCP, Azure
- Containers: Docker, Kubernetes
| Category | Tools / Frameworks |
|---|---|
| AI & LLMs | GPT, BERT, Titan, Hugging Face Transformers, LLM Fine-tuning, RAG Pipelines |
| Deployment | Flask, Streamlit, FastAPI, BentoML |
| Visualization | Tableau, PowerBI, Plotly, Seaborn, Grafana |
| Dev Tools | Git, GitHub, Docker, Postman, MLflow, DVC, DAGsHub, Selenium |
| Collaboration | Jira, SCRUM, Agile |
| Domains | ML, DL, NLP, CV, Generative AI, LLMs, Federated Learning, MLOps, LLMOps, Healthcare AI |
If you find my projects helpful or interesting:
- ⭐ Star my repositories
- 🔔 Follow me on GitHub for updates
- 🤝 Collaborate on AI/ML research and projects
- 💬 Share with others who might benefit
💭 "Render thy labor as sacred offering unto the Most High; and lo, thy toil shall cease to be burden, becoming instead eternal joy"
⭐ If you find my work interesting, consider starring my repositories!
🤝 Open to collaborations, research opportunities, and roles in ML/AI/Healthcare
Made with ❤️ by Divyansh Pandey
💼 Seeking opportunities in world of applied ML/NLP/CV research | 🌐 Let's build something amazing together!