Open to Data / Backend / ML Platform roles

Rohit
Karumanchi.

$ software engineer — data · backend · AI

I build large-scale data platforms and AI systems that ship — from Kafka pipelines moving millions of events to agents that reason over them.

rohit@dev — zsh
~ whoami

rohit — software engineer (data · backend · AI)

~ current --focus

[ kafka-pipelines, langgraph-agents, low-bit-llms ]

~ uptime --career

shipping @ neiman-marcus-group since 2023 · load avg: healthy

~

I'm an AI-focused software engineer who enjoys the deep end: distributed systems, data engineering, and applied ML. Right now I design and deploy large-scale, high-performance data and AI systems at Neiman Marcus Group, supporting both customer-facing analytics and internal decision-making platforms.

Before that, I worked at NNR Global Logistics, tackling large-scale data challenges and modernizing pipelines across critical workflows — where I learned that reliability is a feature users feel.

I hold a Master's in Software Engineering from the University of Texas at Dallas, and I'm driven to explore what's next in data engineering and AI — whether building with founders, collaborating with domain experts, or shipping products that scale. Let's build something that matters.

Neiman Marcus Group

Software Engineer (Full Stack + Data)

2023 — Present

I design and build large-scale, high-performance data and AI-adjacent systems supporting customer-facing analytics, operational intelligence, and internal decision-making platforms. My work sits at the intersection of distributed systems, data engineering, and applied ML.

$1.5M+
Shipping Savings
Real-time Kafka streaming pipelines optimizing carrier selection & route efficiency
250%
Efficiency Gain
Spark/Databricks automation replacing manual ETL with orchestrated data pipelines
$5M+
Annual Cost Savings
ML-powered fraud detection & predictive analytics reducing chargebacks & losses
  • Design and build large-scale data and AI-adjacent systems supporting customer-facing analytics and operational intelligence
  • Translate complex business requirements into reliable, production-grade solutions operating at enterprise scale
  • Partner with stakeholders across business units to translate requirements into scalable, production-grade data products
  • Power downstream analytics for fraud detection, demand forecasting, and inventory optimization using ML models
JavaSpring BootKafkaReact NativeKubernetesAWSTerraformPythonTensorFlowSageMaker

NNR Global Logistics

Software / Data Engineering · prior

Tackled large-scale data challenges and contributed to modernizing pipelines across critical logistics workflows — sharpening the fundamentals of building reliable, scalable systems.

UT Dallas

University of Texas at Dallas

Master's in Software Engineering

Graduate Degree

Featured Build

Nava — High-Performance Dating Platform

Distributed systems architecture for real-time social connections at scale

In Progress — App Store Launch Pending

Repository
10K+
Concurrent Users
<50ms
P99 Latency
99.9%
Match Accuracy
  • Rust/Axum backend engineered for 10K+ concurrent connections per node using Tokio async runtime, stateless horizontal scaling with Redis shared state, and SQLx connection pooling with PgBouncer patterns.
  • GraphQL API layer with async-graphql DataLoader (N+1 elimination), field-level auth, JSONB flexible schemas, composite indexes, and read replica routing for query optimization.
  • Real-time messaging via WebSocket pub/sub with broadcast channels, typing indicators, and read receipts. WebRTC signaling server for video/voice with ICE candidate relay and session state management.
  • ML pipeline using ONNX Runtime (tract) for face detection, liveness verification, and embedding extraction. Vector similarity search with cosine distance for fraud detection and duplicate account prevention.
  • Intelligent matching engine with multi-dimensional compatibility scoring, geo-filtering, bloom filters for seen-user deduplication, and configurable ranking weights for AI-assisted discovery.
  • Production infrastructure: Prometheus metrics (latency histograms, error budgets), distributed tracing with correlation IDs, Kubernetes on Docker multi-stage builds, and MCP server for analytics.
Rust/Axumasync-graphqlWebSocketWebRTCONNX RuntimeTokioSQLx/PostgresRedisPrometheusDockerKubernetesReact NativePyTorchOpenCVVector DBMCP Server

Supply Chain & Inventory Analytics Platform

Event-driven inventory tracking with Kafka consumer groups, offset management, and predictive ML models

50+
Event Topics
99.99%
Data Accuracy
15%
Stockout Reduction
  • Kafka consumers with manual offset commits, partition rebalancing handlers, and dead letter queues across 50+ inventory event topics.
  • Exactly-once semantics using Kafka transactions and idempotent producers for critical inventory adjustments across distribution centers.
  • ARIMA and Prophet forecasting models for demand prediction, integrated with Airflow DAGs for daily retraining and feature store updates in Snowflake.
  • dbt transformation layer with incremental models, SCD Type 2 snapshots, and data quality tests with Great Expectations.
KafkaKafka StreamsSparkSnowflakedbtAirflowARIMA/ProphetPythonSQL

AI-Powered Order Fulfillment Platform

Python microservices with FastAPI, LangGraph agents, and MCP Server for intelligent order orchestration

40%
Faster Routing
85%
Auto-Resolution
<100ms
API Latency
  • FastAPI microservices for order lifecycle management on DynamoDB, with GSI query patterns and Streams CDC to downstream analytics.
  • LangGraph multi-agent workflows for intelligent order routing, exception handling, and customer communication with human-in-the-loop checkpoints.
  • LangChain RAG pipelines with Pinecone vector stores for order history retrieval and context-aware support automation.
  • MCP Server integration exposing inventory APIs, shipping calculators, and order status endpoints as callable tools for AI agents.
PythonFastAPILangGraphLangChainMCP ServerDynamoDBPineconeLangSmithPrometheusKubernetes

Interactive 3D Portfolio with AI Chatbot

This site — Next.js, Three.js, and an OpenAI-powered assistant on AWS

Source
<3s
Load Time
100
Lighthouse
24/7
Uptime
  • Next.js 16 with React Three Fiber 3D graphics, mobile-first responsive design, and load times under 3 seconds.
  • OpenAI-powered chatbot (Spuff) with streaming responses, plus a validated contact pipeline via Resend.
  • AWS deployment with CloudFront CDN, Route 53, SSL, and CI/CD via GitHub Actions — hybrid SSR/CSR rendering for SEO.
Next.jsTypeScriptThree.jsReact Three FiberOpenAI APITailwind CSSAWSCloudFrontGitHub Actions

Amaravati — Smart City Mobility Platform

Amaravati, Andhra Pradesh

Coming Soon

Data Engineering

KafkaSparkPySparkDatabricksAirflowSnowflakedbtDataFusionRay/AnyscaleKinesisPostgreSQLMongoDBRedisCassandraDynamoDBStream ProcessingData WarehousingSchema DesignQuery Optimization

Full Stack

PythonJavaRustC#ReactTypeScriptNext.jsFastAPIGraphQLReact Native/ExpogRPCRESTASP.NET CoreAzure Functions

AI/ML & Agents

PyTorchTensorFlowLangChainLangGraphMCP ServersFAISSVector DBsRAGFederated LearningRLOn-Device MLAWS BedrockSageMakerOpenCVRecSys

Cloud & Infrastructure

AWSAzureKubernetesDockerTerraformNginxCloudFrontRoute 53GitHub ActionsCI/CDPrometheus/GrafanaMicroservices

3D & Frontend Craft

Three.jsReact Three FiberWebGLTailwind CSSResponsive DesignAnimations

Analytics & Collaboration

Power BITableauStreamlitJiraAgile/ScrumConfluence

Let's build something that matters.

I'm eager to make meaningful impact across healthcare, fintech, social good — or the next big thing. My inbox is always open.

or email directly

Designed & built by Rohit Karumanchi

Next.js × Three.js × Tailwind — grown slowly · shipped ripe