TEERTH CHAUHAN

I’m a final-year B.Tech (Artificial Intelligence) engineer focused on building production-ready AI systems—especially agentic LLM workflows, RAG applications, and computer-vision pipelines. Through internships, I’ve worked end-to-end from model design and fine-tuning to deployment-oriented engineering with measurable results.

At Toothlens, I built an “AI Employee” Claude Code–native agentic ML-engineering harness running on a subscription model with no API keys. I also engineered safety and reliability features, including an allow/ask/deny permission system and a destructive-operation safety hook (6/6 tests passing), plus a research sub-agent and automated per-session archiving.

I developed a dental-photo person-verification pipeline using DINOv2 with LoRA fine-tuning and contrastive learning, paired with a trained YOLO detector (mAP@50: 0.92). On AWS EC2 (Tesla T4), I achieved ROC-AUC 0.934 vs 0.811 off-the-shelf (+0.12) at a 0.24% false-accept rate, and I integrated automation using 4 production n8n workflows with centralized error handling and Slack/Google Chat alerts.

I’ve also built interpretable and guarded AI: I fine-tuned EfficientNet-B4 for 5-class dental-image angle classification (98.9% test accuracy) and added Grad-CAM explainability, and I delivered a dental-plan RAG chatbot with PII and toxicity guardrails. Earlier roles strengthened my fundamentals in real-time scene transformation, 3D mesh reconstruction (Open3D), object detection (YOLOv9), and interactive ML apps using Streamlit.