Rramkumar.dev
Available for opportunities

Ramkumar M

>

11+ years building AI-powered products at scale. From computer vision pipelines across 1,000+ retail stores to launching world-class self-checkout platforms at major US sports venues, designing custom programming languages, and orchestrating multi-agent AI systems.

LinkedInGitHub
System Metrics
Experience
11+years
Stores Deployed
1,000+live
Team Size
25+engineers
Shipping Boost
30%faster

About

whoami

Results-driven engineering leader with over 11 years of experience in AI and machine learning, specializing in developing scalable, cloud-based solutions. Proven track record in product management and leading cross-functional teams to drive innovation and operational efficiency.

Currently serving as Senior Manager of Product & Engineering and Country Head at RadiusAI, overseeing a 25-member team building AI-driven checkout systems deployed across some of the largest venues in the United States.

Passionate about leveraging advanced technologies -- from computer vision and deep learning to custom programming languages and multi-agent orchestration -- to develop products that solve complex challenges in real-world applications.

Current Role
Senior Manager, Product & Engineering
Technical Architect / Country Head
RadiusAIBengaluru, India
Education
B.Tech -- SASTRA University
2010 -- 2014

Showcase

signature work

Led development of a full computer vision video analytics pipeline deployed across 1,000+ retail stores, processing terabytes of daily video data with real-time inference on edge devices.

Real-time people tracking with bounding boxes and track IDs in a retail store

--Real-time multi-person tracking with bbox detection, segmentation masks, and persistent track IDs across frames

▸Managed end-to-end ML Ops lifecycle from dataset curation to production model deployment
▸Developed scalable automatic data annotation pipeline processing terabytes of data daily
▸Scaled model training pipelines to optimise usage of H100 on-prem high-end GPUs
▸Trained deep learning models for person re-id, gesture recognition, and real-time tracking
▸Engineered multi-modal dataset management using LanceDB for large-scale training workloads
▸Designed architecture for real-time inference pipelines on edge machines
PyTorchTensorRTCUDALanceDBDockerKubernetesGCP

Smart Checkout POS

Vision-based self-checkout launched at top US sports venues

Product Launch
0
Avg Checkout Time
Lightning-fast vision-based scanning
0
Sales Boost
Revenue increase at major venue
0
Accuracy Rate
Item identification precision
0
Daily Transactions
Peak volume at sports venues
AI-powered self-checkout kiosk at a stadium concession with computer vision item detection
Vision AI Active
checkout time
<10s

--AI-powered self checkout system: overhead cameras identify items instantly, no barcodes needed

Smart checkout POS dashboard showing real-time item detection, receipt building, and analytics

--Real-time POS dashboard: live camera feed with detection overlays, auto-generated receipt, and performance analytics

Checkout Flow4-step
1
Place Items
Customer places items on scanning surface
2
Vision AI Scan
Multi-angle cameras detect & classify instantly
3
Cart Built
Cart auto-populates in under 1 second
4
Tap & Go
One-tap payment, receipt sent digitally
Deployment Sites
Major US Sports VenueStadium
One of the biggest US sports venues
8K+/game day
National Convenience ChainRetail
1,000+ store deployment
Daily high-volume
Architecture
Vision:Multi-angle cameras, real-time inference, item classification
POS Integration:Payment providers, receipt generation, tax engines
Infrastructure:Kafka event streaming, Redis caching, edge computing
Scale:CI/CD across 1,000+ nodes, 25-member cross-functional team
Computer VisionNode.jsKafkaRedisPostgreSQLAWSDockerPOS Systems
Warp-Speed Checkout

Achieved sub-10-second checkout experiences by eliminating barcode scanning entirely through overhead computer vision.

30% Revenue Uplift

Boosted client sales by 30% at a major sports venue through faster throughput and reduced queue abandonment.

Go-To-Market Leadership

Led 25-member cross-functional team and spearheaded GTM strategy for launch at two of the biggest US sports venues.

Low-Code Testing Engine

Solo-built SaaS platform at Zoho Corp -- drag, drop, and ship tests 30% faster

Platform
50%
More Manageable Tests
30%
Faster Shipping
1
Solo Developer
3
Platforms (Web, iOS, Android)
Drag & Drop Builder

Low-code builder mode where QA engineers drag test actions to compose flows. Auto-generates executable code from visual blocks -- no scripting required.

Custom DSL Language

Purpose-built domain-specific language with a custom parser and AST. Simplified syntax familiar to Java/JS developers, targeting Web, iOS, and Android from one codebase.

Cloud Execution Engine

Distributed runtime that executes test suites in parallel across cloud and local agents. One-click scheduling with real-time pass/fail reporting and screenshot capture.

Architecturesolo build
Frontend:Low-code drag & drop builder, pro-code editor with syntax highlighting, web recorder for action capture
Language:Custom parser, AST generation, grammar designed for intuitive QA workflows with Java/JS-like syntax
Runtime:Distributed execution engine running parallel test suites across cloud nodes with Selenium & Appium
Infrastructure:MEAN stack backend, cloud + local agent execution, CI/CD integration, cross-platform test management
JavaSeleniumAppiumMEAN StackCustom ParserASTDistributed RuntimeCloud AgentsCodeMirror

Agentic AI

multi-agent systems

Deep experience in LangGraph multi-agent orchestration. Designing supervisor-planner-executor patterns with human-in-the-loop, evaluation pipelines, and advanced RAG strategies.

Agent Orchestration GraphClick nodes to explore
orchestration
planning
execution
evaluation
Supervisororchestration

LangGraph multi-agent orchestrator. Routes tasks, manages state, enforces policies.

Core Research & Patterns
ReAct Loop
Iterative Reason + Act cycle for multi-step agent reasoning with tool use
FlowSearcher
Structured search flow with dynamic query reformulation and re-ranking
LLM-as-Judge
Using LLMs for automated evaluation, scoring, and quality assurance of outputs
Page Indexing
Efficient document chunking and hierarchical indexing for fast RAG retrieval
Embedding Search
Vector similarity search across Qdrant, Pinecone, and LanceDB backends
Multi-Agent Orchestration
LangGraph-based supervisor pattern with state management and routing
Vector Stores
QdrantPineconeLanceDB
Frameworks
LangGraphLangChainCrewAIAG-UI

Career

experience
2023 -- Present
Senior Manager -- Product & Engineering
Technical Architect / Country Head
RadiusAIBengaluru
▸Spearheaded Smart Checkout POS boosting client sales by 30%
▸Leading Agentic AI POCs -- Data Intelligence Agent for auto-generated reports
▸Engineered multi-modal dataset management with LanceDB on private infra
▸CI/CD deployment across thousands of nodes
PythonPyTorchNode.jsKafkaKubernetesAWS
2019 -- 2023
Lead Data Scientist
RadiusAIBengaluru
▸Developed automatic data annotation pipeline processing TBs daily
▸Scaled model training on H100 on-prem GPUs
▸Built real-time visual tracking for video analytics pipeline
▸Optimised retail convenience store chain operations
PyTorchTensorRTDockerGCPRedisPostgreSQL
2019
Senior Machine Learning Engineer
Pluto7Bengaluru
▸Published Demand Forecasting pipeline to Google AI Hub
▸Generated potential leads through Google AI Hub publication
GCPPythonTensorFlowFull-Stack
2018 -- 2019
Senior Software Developer
Maximl LabsBengaluru
▸CI/CD processes saving millions in oil refinery shutdown downtime
▸Optimised PWA for offline-first field service management
▸Built custom hybrid cloud deployment solution
Node.jsPWADockerCloud
2014 -- 2018
Machine Learning Engineer
Zoho CorpChennai
▸Designed low-code automation SaaS enabling 50% more manageable test cases
▸Wrote custom DSL with parser and distributed runtime for E2E testing
▸Extended platform to support Web + Mobile from single language
▸Developed icon suggestion engine using advanced algorithms
JavaSeleniumMEAN StackParser DesignDistributed Systems

Skills

tech stack
AI / ML
Deep LearningPyTorchTensorRTMLOpsComputer VisionLangGraphAgentsRAG
Languages
PythonNode.jsJavaRustTypeScriptSelenium
Infrastructure
DockerKubernetesKafkaRedisTemporal IOMicroservices
Cloud & Data
AWSGCPAzurePostgreSQLMongoDBSnowflakeCloudflare
Open to connect

Let's Build Something
Extraordinary

Whether you need an AI product leader, a systems architect, or someone who can design both the model and the language it runs on -- I would love to connect.

LinkedInGitHubEmail
B.Tech -- SASTRA University, India (2010-2014)