Harsh Singh

Ecosystem for AI Models Lifecycle Management

• Led the design and development of a scalable ecosystem for end-to-end AI model lifecycle management
• Created organisation-wide standards for coding, model packaging, testing, and production rollout
• Optimised the core system infrastructure for reliability and performance, delivering the minimum viable product within three months
• Engineered data collection and ingestion pipelines for preprocessing and model training
• Architected bulk model validation pipelines and scalable model training systems

Ecosystem for AI Models Deployment & Inference

• Designed a scalable MLOps ecosystem for deploying, inferencing, and documenting over 50 models
• Owned 60% of the codebase, developing end-to-end pipelines for model deployment
• Engineered CV algorithms including object detection, instance segmentation, face recognition, pose estimation, virtual try-on, lip-sync, video enhancement, and real time skin diagnosis
• Developed NLP solutions such as RVC, TTS, speech recognition, sentiment analysis, and LLM-based applications

Skin Diagnosis Application for Kaya Clinic

• Led development of AI-driven skin diagnosis application analysing patient images
• Trained models for skin condition analysis, improving accuracy through data cleaning
• Built end-to-end MLOps pipelines, enhancing system reliability
• Collaborated with dermatologists to ensure clinical relevance

Automated e-commerce data extraction using Python and Javascript, reducing manual effort by 70%

Created reports featuring hierarchical structures, functional analysis, and visual data representation

Monitored crawler/scraper logs to ensure data accuracy