Nafis Neehal
PhD Candidate (CS) | NLP/LLM (Dataset, Benchmarking, Fine-Tuning, Deploy, Evaluate, RAG) | Applied ML
About Me
As a PhD candidate in Computer Science at Rensselaer Polytechnic Institute, I have been solving complex challenges in healthcare—one of the most data-intensive industries—using advanced AI and machine learning techniques. With over 7 years of experience in applied machine learning, deep learning, and data science, I specialize in developing, fine-tuning, and deploying large-scale models that drive real-world impact.
My current work, particularly through collaborations with IBM, focuses on hybrid LLM architectures, RAG pipelines, and agentic systems, ensuring trustworthiness and mitigating hallucinations in AI-driven decision-making. My skills are transferable across industries.
Research Focus
- LLMs in Clinical Trials: Fine-tuning pre-trained state-of-the-art LLMs to optimize trial protocols and designs.
- LLM Benchmarking: Spearheading the development of CTBench, a benchmark suite for evaluating LLMs in clinical contexts, ensuring reliability and precision in real-world applications.
- Hybrid LLM Architectures: Developing RAG and agent-based approaches to better understand clinical texts and improve Q/A systems with VectorDB integrations.
- Trustworthy LLMs: Ensuring reliability, safety, and mitigation of hallucinations in LLMs through RAG + Dynamic Prompting with guardrails and model validation techniques for high-stakes applications.
- Clinical Trial Equity: Implementing state-of-the-art ML techniques to reduce patient recruitment costs by 25% and ensure trial equity, as seen in the award-winning FRESCA framework.
Technical Expertise
- Languages: Python, R, SQL, C++
- Machine Learning Libraries: Langchain, LlamaIndex, HuggingFace, ChromaDB, Tensorflow, PyTorch, Scikit-Learn, AutoML, Unsloth, Autotrain, PandasAI, Pytorch Lightning, Pytorch DDP, OpenAI, Mistral, GROQ
- Cloud Technologies: AWS (SageMaker, Lambda, EC2), Google Firestore (NoSQL)
- Specialization in LLMs:
- Prompt Engineering (Zero/Few Shot)
- Fine-tuning (PEFT)
- End-to-end RAG Pipeline (Embedding, Ingestion, Indexing, Storing, Query Engines)
- Multi-Agent Frameworks
- Quantization
- Deployment (WebUI + Cloud Serving)
- MLOps: Experiment tracking, Data versioning, Model Registry, Recipes, Containerization, CI/CD (GitHub Actions)
Beyond Research
I enjoy developing creative AI-driven projects, such as fine-tuning LLMs to mimic fictional personalities (e.g., the Chandler Bot) and building innovative voice-based AI applications. My passion lies in advancing the frontiers of AI to create practical solutions that enhance both system efficiency and outcomes, regardless of the industry.