AI engineer shipping agentic systems, production ML, and practical software for real-world operations.

MSc Artificial Intelligence with Distinction. Current work spans agentic evaluation workflows, production computer vision, and software built for environments where latency, reliability, and operator trust matter.

• Agentic AI workflows and evaluation systems for multi-step reasoning tasks
• Production ML and computer vision pipelines with measurable field performance
• Real-time desktop AI experiences with tight latency and privacy constraints
• Offline-first business software built for daily operational use

Real-time retail loss-prevention system designed for live RTSP feeds and low false-alarm operations.

• 0.941 ROC-AUC
• 93.3% accuracy on real footage
• 2.7% false alarm rate
• Multi-stage inference: motion gate, fast pose scoring, asynchronous video models, operator feedback loop

Offline-first workshop operations platform used in live two-wheeler servicing workflows.

• Daily workflow support for invoicing, stock, mechanics, and reminders
• Fast mobile-first UI for shop-floor usage
• Local-first interaction model with cloud sync where needed

Live SaaS product at mapinion.in where local businesses can subscribe, buy plans, and turn one printed QR into a guided review-growth system.

• Guided customer flow with editable review suggestions
• Fixed QR architecture that survives business preference changes
• Owner dashboard and operational controls for review inventory
• Paid plans and direct website purchase flow for business customers

Low-latency desktop AI assistant for live spoken workflows, contextual suggestions, and privacy-aware overlay design.

• Sub-second response flow
• Real-time audio capture and transcript-to-suggestion pipeline
• Context injection and on-screen assistance for high-pressure conversations
• Overlay/privacy mechanics engineered for user control and separation from meeting content

• Built production systems across computer vision, business operations, and real-time AI assistance
• Shipped software for live environments where uptime, speed, and operator trust are non-negotiable
• Comfortable across model behavior, backend systems, and product-facing workflow design

• Agentic workflows
• Evaluation and human feedback systems
• Applied ML infrastructure
• Low-latency AI product experiences

• AI: internal prototyping around AI tooling and applied workflows
• Bet: product and growth experiments kept secondary to the main engineering story

• Email: 1.priyannsh@gmail.com
• LinkedIn: linkedin.com/in/priyansh-1221
• Portfolio: priyansh1221.github.io/Priyansh