Ritvik
Reddy
Gangula
Software engineer building backend systems that stay up when parts of them fail, and the AI agents that run on top of them. Computer science at Arizona State, graduating May 2027.
Reach me
Now exploring
Oct 2026
- Evaluating LLM systems
- LLM as a judge
- RAG and ways to improve retrieval
About
I'm a software engineer who likes the layer underneath the product: queues, retries, and the data paths that decide whether a system holds up.
- School
- Arizona State University
- Degree
- B.S. Computer Science, May 2027
- GPA
- 4.0, Dean's List six times
- Certified
- AWS Cloud Practitioner
Experience
Jun – Aug 2026
, Chicago, ILA $96B AUM investment firm. I built a RAG service on Azure OpenAI and backend services for client-facing platforms.
- 560+ team hours saved per quarter
- 730+ clients served
- 1.5s median retrieval
- Engineered a RAG service in C#/.NET on Azure OpenAI (GPT-4o) that drafts client reports grounded in research documents via hybrid retrieval (Azure AI Search, 1.5s median retrieval latency), covering 500+ portfolios and saving 560+ team hours per quarter
- Built production backend services for an in-house document portal replacing a third-party platform, using ASP.NET Core Web API, EF Core, and SQL Server, with client-scoped RBAC and SAS-based downloads from Azure Blob Storage, serving 730+ clients and 500K+ migrated documents, tested with xUnit and Cypress E2E
- Implemented distributed tracing, log correlation, and production dashboards across 6+ .NET web apps and Azure Functions with Datadog APM and Sumo Logic (Serilog trace IDs), cutting diagnosis of a production Azure Function timeout from hours to under 30 minutes
C#/.NET / ASP.NET Core / EF Core / SQL Server / Azure OpenAI / Azure AI Search / Azure Functions / Datadog / xUnit / Cypress
Visit GCM GrosvenorJune – May 2025
, Philadelphia, PAAutomated data enrichment and quality validation pipelines using Python and GPT, improving dataset reliability across analytics workflows.
- 25+ hours saved per week
- ~40% more accurate data
- Improved dataset accuracy by ~40% through automated validation and consistency checks
- Reduced manual data processing by 35%+ by streamlining Google Sheets–based workflows
- Implemented anomaly detection to identify data inconsistencies across production datasets
- Strengthened downstream ML and analytics reliability with preventative data checks
Python / OpenAI API / Google Sheets API / Data analysis
Visit Food Forest AIMay – Aug 2024
, Hyderabad, IndiaBackend APIs and a fast React frontend for a civic permit portal used by 15,000+ citizens and officers.
- 75% faster reads
- 3.4s → 0.8s render
- 65% fewer search calls
- Built cursor-based pagination on DynamoDB (LastEvaluatedKey) and a React frontend with debounced search and react-window virtualization, cutting search API calls by 65% and initial render of 5K+ record permit histories from 3.4s to 0.8s
- Replaced full-table scans on a 50K+ record permit table with key-based queries by redesigning the DynamoDB partition and sort keys and adding a status GSI, cutting average read latency 75% (400ms to 100ms, CloudWatch)
- Built 15+ REST APIs in Node.js/Express (TypeScript) with Joi request validation and Okta-based authentication and RBAC, enforcing 6 permission tiers through authorization middleware and role-based guards
Node.js / Express / TypeScript / React / DynamoDB / AWS / Okta / Joi
Projects
August 2026
A three-agent LangGraph pipeline that turns SEC EDGAR filings into classified, cited findings, served as an MCP server and REST API on AWS.
- 3-agent LangGraph pipeline
- 90% eval gate in CI
- MCP server on AWS
- Aligner, Classifier, and Verifier agents; the Verifier retries retrieval once and drops unsupported findings
- Gated in CI by a 30-pair hand-labeled eval set at a 90% accuracy threshold
- Deployed as an MCP server and REST API with SQS-queued async requests, response caching, and Step Functions orchestration
- All infrastructure provisioned with Terraform
Python / LangGraph / MCP / AWS Lambda / API Gateway / SQS / Step Functions / Terraform
January 2026
A distributed job orchestrator in Go with Raft-replicated coordinators, gRPC workers, and a Kafka-backed event log.
- 16.7 jobs/s
- 145 ms leader failover
- 0 acknowledged jobs lost
- 3 Raft-replicated coordinators and 25 gRPC workers sustaining 16.7 jobs/s across 1,000 real job submissions
- Recovered every acknowledged job across 5 process-kill trials, with a 145 ms median leader failover
- Followers forward writes to the leader; job state is durably replicated via Kafka
- Prometheus/Grafana dashboards plus automated crash and partition fault-injection tests
Go / gRPC / Kafka / Raft / k3s / Prometheus / Grafana
June 2025
A full-stack AI system that decomposes topics into parallel sub-questions and synthesizes source-backed research reports using a stateful agent workflow.
- Parallel LangGraph agents
- Live agent graph UI
- Pinecone memory
- Orchestrates planner, researcher, aggregator, critic, and synthesizer agents using LangGraph with parallel execution
- Streams real-time node-level updates to a React UI with live agent graph visualization
- Uses Pinecone vector memory to retrieve past context and improve future research quality
Gradio / CrewAI / LangGraph / Python / AI Agents
December 2024
Real-time chat application with rooms, presence indicators, and multi-device conversation sync.
- Real-time rooms
- JWT + Redis rate limits
- Dockerized
March 2026
AI-powered platform for intelligent meeting transcript analysis with hierarchical NLP summarization and action item extraction.
- BERT semantic chunking
- Action item extraction
- FastAPI + Next.js
- BERT-based semantic chunking with overlap for context preservation
- Hierarchical extractive summarization (chunk → overall)
- Pattern-based action item extraction with assignee & deadline detection
- Secure JWT auth with bcrypt and per-user data isolation
FastAPI / Next.js / Python / SQLAlchemy / BERT / Hugging Face / JWT / TypeScript
May 2025
AI-powered career assistant providing personalized guidance and recommendations.
- LangChain
- Hugging Face models
- Deployed on Vercel
- Personalized recommendations
- Natural language processing
- Cloud deployment
LangChain / Hugging Face / Vercel / React / AI/ML
December 2024
LSTM-based machine learning model for accurate stock price forecasting.
- LSTM model
- TensorFlow / Keras
- Time series
- High prediction accuracy
- Time series analysis
- Data visualization
TensorFlow / Keras / LSTM / Pandas / Matplotlib / Python
Skills
Languages
- Go
- Python
- Java
- C#
- TypeScript
- JavaScript
- SQL
Backend & frameworks
- .NET / ASP.NET Core
- EF Core
- Node.js / Express
- React
- Angular
- REST APIs
- gRPC
- JWT / Okta
Cloud & DevOps
- AWS Lambda, SQS, Step Functions
- Azure OpenAI, AI Search, Functions
- Terraform
- Docker
- Kubernetes (k3s)
- GitHub Actions
Distributed systems & data
- Kafka
- Raft (hashicorp/raft)
- SQL Server
- DynamoDB
- MongoDB
- Redis
Agentic AI & ML
- LangGraph
- MCP
- RAG
- LangChain
- CrewAI
- TensorFlow
- Hugging Face
Observability & testing
- Datadog APM
- Prometheus
- Grafana
- Sumo Logic
- xUnit
- Cypress
Get in touch
I'm always open to discussing new opportunities, interesting projects, or just having a chat about technology.