jignesh mandana

I’m an AI/ML engineer focused on turning real-world problems into production-ready models and reliable ML pipelines. From detecting driver fatigue on Indian roads to automating radiology triage in under-resourced hospitals, I care about measurable impact, not just accuracy.

As the founder of an AI agency, I delivered 10+ production AI solutions for clients across the US, UK, and Canada, maintaining a 95% client retention rate. I also built LLM-powered CRM automation pipelines that eliminated 60% of manual operations, saving 10+ hours per week per client.

My technical edge is end-to-end ML: fine-tuning LLMs and vision-language models with LoRA/DoRA using HuggingFace Transformers and Unsloth, then deploying real-time inference on serverless GPU infrastructure (Modal, HuggingFace Spaces) with Weights & Biases experiment tracking. I like pipelines that are fast, cost-aware, and easy to trust.

Highlights include a Gemma 4 VLM fatigue scoring system powered by camera processing and MediaPipe FaceMesh running on a single T4 GPU at zero marginal cost, a YOLOv8 road damage workflow that slashes review time by ~94%, and CheXNet-style radiology screening with Grad-CAM interpretability. I’m currently competing in the Kaggle Gemma 4 Impact Challenge ($200K) and looking for full-time AI/ML Engineer and MLOps Engineer roles.