Sanskar Pandey

Pioneered a real-time analytics dashboard using Next.js, TimescaleDB, and Recharts/ApexCharts/ECharts, enabling 3D product and sales insights across 10K+ SKUs with millisecond-level responsiveness. Engineered a scalable data ingestion pipeline with FastAPI, Redis, and TimescaleDB, processing over 500K+ daily events with schema validation, batching, and UPSERT logic. Deployed full observability stack (Prometheus, Grafana, Loki, Slack alerts) via Docker, reducing average incident resolution time by 60%. Developed a custom 360° Viewer JS library to animate high-res product embeds, improving user interaction rates by 30% on digitized catalog pages.

Innovated 2 SaaS solutions, an AI-powered document storage and management system and an online exam platform using Django on AWS, achieving 95%+ processing success and API responses of sub 2 seconds. Designed and optimized Elasticsearch-based search pipelines with custom analyzers and inverted indexes for fast, scalable information retrieval across large corpora. Streamlined secure authentication (Google OAuth, JWT) and payment integration (Razorpay), increasing user engagement by 30%.

Revamped a survey platform by migrating the backend from Flask to FastAPI and integrating a Next.js frontend, reducing API latency by 40% and maintaining sub 2 second query times. Engineered end-to-end cloud deployments on Azure and AWS with Docker, Kubernetes, and Nginx on Linux VMs; tuned kernel TCP/IP parameters to increase throughput by 30% and ensure 99.9% uptime. Integrated advanced AI analytics with Pinecone using an ensemble model (Azure OpenAI, Claude), boosting NLP accuracy by 25% and facilitating real-time data insights. Architected a high-availability job queue using BullMQ and Redis across 10 distributed worker nodes, ensuring zero task loss under 500+ RPS.

Contributed to and maintained open-source Python repositories, enhancing test coverage, refining CI workflows, and ensuring production-grade standards across all code bases. Designed robust data pipelines to generate and validate LLM training datasets, integrating custom validators and schema enforcement for high-quality, industrial grade output. Built scalable, automated workflows for dataset creation using Python, reducing manual overhead by 70% and accelerating training readiness for multiple LLM iterations.