8000 machhakiran (Machha kiran) Β· GitHub
[go: up one dir, main page]

Skip to content
View machhakiran's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report machhakiran

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
machhakiran/README.md

🌐 Executive Profile

I am Kiran Machha, a Forward Deployed AI Engineer based in Singapore, specializing in Sovereign AI Infrastructure, Autonomous Multi-Agent Systems, and Air-Gapped Enterprise AI Platforms.

I bridge the gap between frontier AI research and resilient enterprise operations:

  • πŸ›οΈ Sovereign AI and Air-Gapped Deployments: Packaging multi-model inference stacks (vLLM, llama-server, Ollama) with zero cloud data egress and automated hardware auto-tuning across NVIDIA CUDA, Apple Silicon Metal, and AMD ROCm.
  • πŸ€– Autonomous Agents and Tool Orchestration: Architecting production agent loops around the Model Context Protocol (MCP), OpenClaw, and Hermes Agent harnesses with stateful SQLite memory and self-correcting reflection loops.
  • 🧠 Enterprise Hybrid RAG and Vector Systems: Designing dual-engine retrieval architectures with OpenSearch 2.19 (BM25 keyword matching + dense vector embeddings with Reciprocal Rank Fusion) and sub-2ms Redis semantic caching.
  • πŸ›‘οΈ Enterprise Grounding: Backed by a strong engineering foundation in regulated banking change control, microservices, and Kubernetes GPU node pool management.

πŸ›οΈ Core Engineering Pillars and Production Capabilities

πŸ›οΈ Sovereign AI and Local Inference

  • Air-Gapped AI Platforms: Self-hosted multi-service deployments with 100% data privacy and zero cloud leakage.
  • Hardware Acceleration: Automated hardware overlays for NVIDIA CUDA Tensor Cores, Apple Metal, and AMD ROCm.
  • High-Throughput Model Serving: vLLM, llama-server, Ollama, and LiteLLM proxy gateways.
  • Privacy-Shield DLP: Real-time PII, token, credential, and credit card redaction before prompt dispatch.

πŸ€– Agentic AI and Swarm Protocols (MCP)

  • Model Context Protocol (MCP): Native server and client implementations connecting IDEs to browser tools.
  • Autonomous Agent Engines: OpenClaw (:18789) and Hermes Agent (:9120) with SOUL.md state.
  • Multi-Agent Swarm Pipelines: 3-phase autonomous pipelines: PRD Architecture βž” Swarm Build βž” Playwright Pilot.
  • Stateful Memory Systems: Transparent three-tier SQLite stores (semantic, episodic, and procedural).

🧠 Enterprise Hybrid RAG and Knowledge Systems

  • Reciprocal Rank Fusion (RRF): Merging BM25 keyword precision with dense semantic vectors in OpenSearch 2.19.
  • Sub-2ms Semantic Caching: Redis vector caching delivering 150x–400x latency speedups on repeated queries.
  • Agentic LangGraph Workflows: Document grading, automatic query rewriters, and strict citation enforcement.
  • Vector Databases: Production deployments on Qdrant, PGVector, and OpenSearch.

⚑ Realtime Voice, Vision and GPU Cloud

  • Low-Latency Realtime Voice AI: Sub-800ms full-duplex voice agents with Twilio SIP, FastRTC, Faster-Whisper, and Kokoro TTS.
  • Computer Vision and OCR: Bank cheque OCR with OpenCV deskew, MICR E-13B readers, and automated fraud parity verification.
  • Asynchronous GPU Pipelines: High-concurrency (10k req/s) Rust Axum gateways + Redis task queues + vLLM vision workers.
  • LLMOps and Observability: Full telemetry tracing with Langfuse, Opik, and LLM-as-Judge eval harnesses.

πŸš€ 11 Featured Production Systems Showcase

A selection of real, deployed production repositories and architectures engineered by Kiran Machha. Inspect full interactive architecture topology flows on machhakiran.pro:

# System Name Category Architectural Highlights Core Tech Stack
01 OpenForge AI Sovereign AI Infrastructure 24 bundled containerized services, OpenClaw and Hermes agents, LiteLLM gateway, Qdrant hybrid RAG, hardware auto-overlays (CUDA/Metal/ROCm). Docker, LiteLLM, Qdrant, SearXNG, ComfyUI, vLLM
02 Personal AI Agent (KaviAgent) Autonomous Local Agent Local-first personal assistant with 3-tier SQLite memory, ~95-line plain Python loop, web cockpit (:7777), and Telegram bot bridge. Python 3.12, SQLite, FastAPI, Claude API, Ollama
03 Agentic Workspace IDE (KaviSpace) Desktop Multi-Agent IDE Tauri v2 + Rust native shell with xterm.js terminals and KaviSwarm 3-phase autonomous pipeline (PRD βž” Build βž” Playwright Pilot). Tauri v2 (Rust), React 19, xterm.js, Express, Playwright
04 Business Growth Agents Platform (KaviAI) Multi-Channel AI SaaS Autonomous multi-channel agent squad generating and publishing across 10+ platforms (Reddit, LinkedIn, X, YouTube Shorts). Next.js 16, React 19, Gemini 2.5 Flash, Supabase RLS
05 Agentic Coding Platform and Sandboxes Developer AI Infrastructure Self-hosted ephemeral Docker/K8s sandboxes with Model Context Protocol (MCP), AST tools, secret scanning, and token budgets. FastAPI, Kubernetes, Docker, MCP Protocol, UV
06 Corporate Enterprise RAG Enterprise Hybrid RAG Air-gapped enterprise knowledge system with OpenSearch 2.19 (BM25 + Dense RRF), Airflow 3.0, and LangGraph agentic loop. OpenSearch 2.19, LangGraph, FastAPI, Redis, Ollama
07 Bank Cheque OCR Automation Banking and Financial AI High-volume cheque OCR API (5,000+ daily) with OpenCV deskew, MICR E-13B strip reader, and automated digits-vs-words fraud parity checks. FastAPI, OpenCV, PyTesseract, CRNN, Pydantic v2
08 Customer Call Support Voice Agent Realtime Voice AI Sub-800ms full-duplex voice agent on Twilio SIP, FastRTC low-latency audio transport, Moonshine STT, and Kokoro 3B TTS on GPU pods. FastRTC, Twilio WebRTC/SIP, Qdrant, Superlinked, Opik
09 n8n AI Agents and Workflow Collection Agentic Workflows 17 production enterprise workflows with GPT-4o tool calling, strict JSON schema validation, automated IT ticket triage, and alert routing. n8n Automation, OpenAI GPT-4o, Gmail API, Webhooks
10 Invoice Automation OCR Sovereign GPU Pipeline Multi-stage invoice processing with 10k req/s Rust Axum gateway, Redis async GPU queues, and vLLM vision models (Qwen2-VL). Rust (Axum/Tokio), vLLM Vision, Redis, PostgreSQL
11 PageBolt MCP Server Developer AI Tooling Open-source MCP server providing 9 typed tools for web screenshots, PDF generation, and token-budgeted DOM inspection for AI agents. Model Context Protocol (MCP), TypeScript, Puppeteer

πŸ› οΈ Production Technology Stack






πŸ’Ό Enterprise Track Record

  • Forward Deployed AI Engineer β€” Private AI and Agent Systems (Singapore) Present
    • Deploys sovereign AI systems, multi-agent swarms, and self-correcting reasoning loops with enterprise clients across APAC.
    • Manages the full LLMOps lifecycle: local serving, LiteLLM gateway routing, LLM-as-Judge evaluation harnesses, and Langfuse tracing.
    • Provisions Kubernetes infrastructure with GPU node autoscaling (scale-to-zero during off-peak hours).
  • AI Platform / LLM Ops Engineer β€” Enterprise SaaS
    • Built the shared inference platform with model gateways, usage-based metering, and semantic caching to minimize cost per request.
    • Implemented automated evaluation gates in CI/CD pipelines so prompt and model updates were deployed based on empirical benchmarks.
  • Full Stack AI Engineer β€” Payments and FinTech
    • Delivered end-to-end customer-facing AI services on Next.js and FastAPI with hybrid search and RAG over transaction corpora.
    • Hardened production pathways with asynchronous workers, idempotent retry queues, and structured tracing.
  • Technical Lead Engineer β€” Enterprise Applications β€” Regulated Banking
    • Built and maintained enterprise-grade banking applications under regulated change control.
    • Designed microservices on Spring Boot and Node.js with strict security reviews, access control, and compliant audit logging.

πŸ“Š GitHub Activity and Telemetry


🀝 Connect with Kiran Machha

Available for Forward Deployed AI Engineer, Sovereign AI Infrastructure, and Principal Agentic Architect roles across Singapore, APAC, and EMEA.


"The future isn't just about AI that talks. It's about AI that acts β€” securely, locally, and autonomously."

Pinned Loading

  1. FinBot-AI FinBot-AI Public

    Agentic Financial Intelligence System

    HTML 1

  2. Agents-1 Agents-1 Public

    From Jira to PR. Claude-Powered AI Agents that Code, Test, and Review for You

    Python

  3. SmartDoc-Curator SmartDoc-Curator Public

    SmartDoc Curator is an enterprise-grade Agentic Retrieval-Augmented Generation (RAG) platform built for banking and financial institutions.

    Python

  4. ai-cli-coding-asssit ai-cli-coding-asssit Public

    Python

  5. Agentic_AI_with_LangGraph Agentic_AI_with_LangGraph Public

    Jupyter Notebook 1

  6. ai-infra-engineer-learning ai-infra-engineer-learning Public

    Forked from ai-infra-curriculum/ai-infra-engineer-learning

    AI Infrastructure Engineer Learning Track - Production ML infrastructure curriculum (2-4 years experience)

    Python

0