I am a Master's student in Computer Science at New Jersey Institute of Technology (NJIT), specializing in Natural Language Processing, Generative AI, and Scalable System Design. Bridging the gap between research and production, I have experience architecting full-stack RAG-based search engines, optimizing LLM pipelines, and deploying containerized microservices. My passion lies in building robust AI systems that solve real-world problems in recruitment, healthcare, and security.

Medi-Gemma CDSS | PyTorch, LlamaIndex, LLaVA, Gemma 3, MedGemma, Hulu-Med, Streamlit

• Multimodal CDSS Architecture: Engineered a Clinical Decision Support System combining a fine-tuned Vision-Language Model (LLaVA-Medical-Director) with an Agentic RAG pipeline to analyze wound pathology from clinical images and patient EMR data.
• Deterministic Safety Layer: Built a keyword-driven ProtocolManager that maps VLM diagnoses to a tiered clinical protocol hierarchy, grounding treatment recommendations in auditable rules and preventing LLM hallucinations from reaching the triage stage.
• Advanced Clinical Benchmarking: Developed a rigorous LLM-as-Judge evaluation pipeline (judged by Claude Opus 4.6) testing foundation models against a 9-question benchmark and a strict 7-dimension clinical safety rubric across 63 patient records.
• Key Performance & Ablation Results:
  • MedGemma-1.5 (4B): Achieved a 63/63 benchmark pass rate with a 6.5s load time, using targeted anti-refusal prompt alignment (clinician_v3_mg4b).
  • MedGemma (27B) & Gemma 3: Validated at 0 CRITICAL_FAILs, establishing them as the strongest stability and safety baselines for clinical deployment.
  • Hulu-Med (32B): Reduced MISSED_EMERGENCY triage failures from 3 to 1 by enabling step-by-step thinking decoding (use_think: true).

ScoutIQ | Llama-3, Qdrant, FastAPI, Python

• Developed an Agentic RAG-based search engine and recruitment intelligence platform.
• Optimized LLM pipelines and vector retrieval for high-accuracy semantic search.

• 🔭 I’m currently building ScoutIQ, an AI-powered recruitment intelligence platform leveraging Llama-3, Vector Search (Qdrant), and RAG pipelines.

• 🌱 I’m currently mastering Advanced Retrieval-Augmented Generation (RAG) techniques, Kubernetes orchestration for ML workflows, and Agentic AI patterns.

• 👯 I'm eager to collaborate on projects involving Large Language Models (LLMs), Semantic Search, MLOps, and scalable backend architectures.

• 🤝 I’m open to discussions on optimizing vector retrieval, fine-tuning open-source LLMs, and securing AI supply chains.

• 👨‍💻 All of my projects are available at https://github.com/MohammedSaim-Quadri

• 📫 How to reach me: mohammedsaimquadri@gmail.com | LinkedIn

• ⚡ Fun fact: I enjoy breaking down complex AI topics into simple explanations—sometimes I learn best by teaching!

• How LSTMs Master Long-Term Dependencies
• Artificial Neural Networks (ANN) vs. Recurrent Neural Networks (RNN)
• Word2Vec Decoded: Building Semantic Understanding with CBOW and Skip-grams
• The NLP Foundational Blueprint: From Raw Text to Numerical Vectors
• Forceful Generalization: Mastering Dropout Layers in Deep Networks
• The Flip Side of the Coin: Exploding Gradients and Weight Initialization

Programming Languages

Machine Learning & Data Science

Backend & Web Frameworks

Databases, Cloud & DevOps