I build production backend systems and AI-native products with Python, cloud infrastructure, MCP, RAG, and agentic workflows.
Seven years of Python engineering across SaaS, distributed systems, cloud platforms, and AI applications. I care about the parts that make AI useful in production: reliable APIs, measurable retrieval quality, observability, evaluation, cost control, and safe deployment.
- 516 automated tests in CI for Zocept: 358 Playwright end-to-end tests and 158 Vitest unit tests
- 213 credit cards supported by Zocept, a live AI-powered rewards optimization product
- MCP implementation merged upstream in Oracle AI Optimizer
- Built both sides of an MCP integration: a vector-search MCP server and an MCP-enabled AI Optimizer client/server
- Led architecture for a multi-tenant Azure SaaS platform and mentored a team of 6 engineers
- Reduced downtime incidents by 40% and incident response time by 25% through architecture and observability improvements
- Designed cloud infrastructure across AWS, Azure, GCP, VMware, and OpenStack
- Built real-time data synchronization serving 50,000+ users
- Automated manual infrastructure workflows, reducing processing time by 90%
- Reduced large-scale analytics runtime by 40% with PySpark and Amazon Redshift
- MCP servers and clients
- Retrieval-augmented generation pipelines
- Multi-agent orchestration and verification loops
- Retrieval evaluation and golden datasets
- Local-first AI systems with Ollama and vector databases
- AI application APIs with FastAPI
- Model routing, prompt versioning, and cost-aware workflows
- Python services with FastAPI, Django, and Django REST Framework
- Event-driven and serverless architectures
- Multi-tenant SaaS platforms
- Distributed systems and real-time communication
- PostgreSQL, Redis, MySQL, SQLite, and Redshift
- CI/CD, infrastructure as code, and production observability
AI-powered credit card rewards optimizer for Indian users.
- Built and operate the product independently from MVP to production
- Covers 213 credit cards across major Indian banks
- 516 automated tests run in CI on every deployment
- 358 Playwright end-to-end tests and 158 Vitest unit tests
- Production Lighthouse scores: 90/100/100/100
- FastAPI backend, React frontend, Supabase, Razorpay, and Vercel
Open-source MCP implementation contributed during my work at Oracle.
- Implemented MCP server and client capabilities
- Built a vector-search MCP server
- Exposed the AI Optimizer as an MCP server
- Enabled the Optimizer to consume external MCP capabilities as a client
- Added vector-store documentation, disclaimers, and test coverage
Privacy-first, local-first AI knowledge product.
- Ingests personal data such as chat exports, bookmarks, and email
- Uses Ollama and local vector search
- Works with internet access disabled
- Current hardening includes retrieval evaluation, role-based access control, and agent-level observability
- Architected a multi-tenant e-commerce SaaS on Azure using event-driven serverless patterns
- Reduced downtime incidents by 40%
- Established Grafana Loki observability standards
- Reduced incident response time by 25%
- Mentored a squad of 6 engineers from MVP toward production
- Built SaaS features with Python, Django, DRF, and PostgreSQL
- Improved core system stability by 30%
- Designed infrastructure with Terraform and CloudFormation across multiple cloud and virtualization platforms
- Supported a 25% increase in deployment load capacity
- Partnered directly with product teams to turn ambiguous requirements into maintainable systems
- Automated approximately 50% of manual infrastructure tasks with Python and VB scripts
- Reduced processing time for those workflows by 90%
- Reduced analytics runtime by 40% with PySpark and Amazon Redshift
- Built real-time WebSocket infrastructure for live synchronization across 50,000+ users
Languages: Python, JavaScript, C++, C, PHP, SQL, Bash
AI: MCP, RAG, agentic workflows, multi-agent orchestration, evaluation pipelines, Ollama, LiteLLM, Anthropic, OpenAI, Gemini, Hugging Face
Backend: FastAPI, Django, Django REST Framework, React, PostgreSQL, Redis, MySQL, SQLite, Redshift
Cloud and Infrastructure: AWS, Azure, GCP, Kubernetes, Docker, Terraform, CloudFormation, VMware, OpenStack, CI/CD
Observability and Data: Grafana Loki, PySpark, vector databases, retrieval evaluation, distributed systems
- Senior AI Engineer roles
- Senior Backend Engineer roles
- Forward-deployed and applied AI engineering roles
- MCP, RAG, agentic systems, and AI platform consulting
- Collaborating on practical open-source AI projects
GitHub: [github.com/Ishuin](https://github.com/Ishuin)
LinkedIn: [linkedin.com/in/ishukumars](https://www.linkedin.com/in/ishukumars)
I build AI systems that are useful, observable, testable, and difficult to fool.