TANMAY AGRAWAL

Spearheaded development of a customer onboarding automation flow, reducing onboarding time by ~66% and accelerating client go-live speed.

Led the planning and implementation of a project from scratch - "Automated PDP Generation". Orchestrated things from inception to completion – from writing the tech spec, architecting scalable Temporal.io workflows, distributing tasks, setting CI/CD pipeline, to managing the Kubernetes (GKE) deployments.

Achieved ~70% reduction in client onboarding time and ~50% bandwidth increase for product operations team via Automated PDP Generation initiatives.

Built product clustering system using Agglomerative Hierarchical Clustering (AHC) to process large-scale JSON datasets and automatically group products with similar features, resulting in optimized storefront UX.

Drove ~50% increase in conversion rates in 3 months by collaborating on designing and deploying new experiment-driven shopping experiences.

Developed high-performance retail consumer-facing components using Next.js, Tailwind CSS, and architected the dashboard platform modules with React, Chakra UI, and Apollo GraphQL.

Designed and deployed scalable GraphQL APIs with Hasura; implemented mission-critical backend services (checkout, cart validation, product-info) using Golang.

Led a self-serve signup and payment workflow project, integrating Stripe and Auth0 to implement identity-aware access controls, permission-based account state transitions, and PCI-compliant payment processing.

Contributed to the development of Pierre, the Al agent for platform, integrating OpenAl and Anthropic APIs and optimizing prompt engineering for faster, high-fidelity LLM responses.

Improved platform reliability and developer workflow efficiency by automating production DB migrations with CI/CD pipelines using Flyway & GCP Build trigger.

Engineered a real-time radar data ingestion pipeline with synchronized frequency control, achieving consistent 100ms output periodicity across C-based embedded firmware and Python-based desktop app

Built a low-latency, multithreaded system to decode high-frequency binary device streams and render real-time visualizations for radar point clouds.

Architected a radar GUI application using wxPython for dynamic analysis and visualization of 3D sensor data streams in live environments.

Developed a presence detection algorithm by applying DBSCAN clustering on point-cloud data with Scikit-learn, enhancing device intelligence.

Contributed to research and prototyping of object classification strategies, improving radar system capabilities for future automotive and loT products.

Prototyped classification models using PyTorch

Built REST-based integrations with third-party apps (Google Drive, Jira, Slack), enabling data ingestion pipelines for LLM-based augmentation.

Integrated OpenSearch to store vector embeddings.

Designed and deployed event-driven Temporal.io workflows triggered by MySQL CDC events, orchestrating real-time file/text extraction and semantic indexing via Llamalndex.

Engineered dynamic search and filtering components in React using Tailwind CSS, improving UX responsiveness for large document datasets.

Optimized indexing throughput via API request parallelization, achieving 50x performance gain