I’m Arshad Jafri S MD — a passionate full-stack developer and aspiring ML / data scientist who loves to bridge code and data.

My work spans web development, backend systems, and machine learning, all driven by a mission to build scalable, intelligent, and user-centric applications.

✨ Creating bugs since … 2000 (some might say I was born debugging)
📚 Learning mode: transformers, large language models, biomedical applications and full-stack applications — and yes, I’m also learning how to actually rest
🎲 I love good food, weird jokes, and soundtracks that make me feel like I’m walking through a cinematic story — mysterious, brooding, with a hint of hidden power

• Built predictive models from MRI, fMRI, and DTI scans to stratify PTSD risk — making neuroimaging less mysterious (and more accurate).
• Designed preprocessing pipelines (normalization, augmentation, dimensionality reduction) and cut training time by ~20%.
• Fine-tuned deep architectures (CNNs, RNNs, Transformers) while tuning hyperparameters — because “just guess” rarely hits the mark.
• Added explainability (Grad-CAM, SHAP) to spotlight brain regions correlated with PTSD — turning “black box” into “somewhat transparent box.”
• Automated experiments using MLOps tools (Weights & Biases, Docker, GitHub Actions) — reduced reproducibility errors by ~40%.
• Collaborated with neuroscientists to ensure model findings align with psychiatric protocols — bridging code and clinical insight.

• Transformed structured PTSD datasets into narrative “clinical transcripts” for compatibility with LLMs — avoided format mismatches by ~30%.
• Built NLP pipelines in Hugging Face / PyTorch to process 25,000+ text samples — carefully handling each token.
• Fine-tuned transformer models (Centaur, LLaMA) to achieve smoother convergence and lower validation loss.
• Engineered structured prompts for clinical events — increasing consistency and classification accuracy.
• Automated logging, evaluation, and orchestration using Weights & Biases — speeding up iteration cycles by ~25%.

• Crafted departmental websites using HTML, CSS, JavaScript, and CMS platforms (WordPress, Sitecore) — making slow pages a thing of the past.
• Integrated RESTful APIs and backend services to enable dynamic features — reduced page load times by ~25% across thousands of visits.
• Remediated web content for WCAG / accessibility standards — ensuring users of all abilities can browse comfortably.

• Developed and maintained enterprise insurance applications using Java, Spring Boot, JSP / Servlets — features used by 50,000+ policyholders.
• Built reusable backend components and APIs — speeding up product rollout cycles by ~20%.
• Collaborated with analysts and stakeholders to translate domain requirements into tech solutions — turning “business speak” into “code speak.”
• Diagnosed and debugged complex production issues (logs, monitoring) with ~95%+ resolution success — minimizing downtime.
• Optimized database queries (MySQL) and refactored core modules — improving transaction speeds by ~30%.
• Participated in Agile sprints, peer code reviews, CI/CD (Jenkins / Git), and handled legacy version control (SVN) — reducing release times by ~25%.