ankit.

AI engineering × cloud infrastructure

Useful AI.
Solid foundations.

I’m Ankit Vishwakarma, an AI engineer working with agents, LLMs, and RAG. Currently a Senior Software Engineer at Creative Synergies Group, with a foundation in DevOps and cloud infrastructure.

Explore my work GitHub
Agentic AILLMs & RAGMLOpsCloudOps
Built in the open. Shared on GitHub.

01 / Selected work

From idea to implementation.

A focused selection of AI projects: agents, retrieval, local inference, generative imaging, and context-aware applications.

Featured project / agent memory

RAG-MCP

A memory layer beyond the chat.

Persistent, searchable knowledge for MCP-compatible assistants. Bring together documents, web pages, notes, and YouTube transcripts — then retrieve relevant context with source citations and namespace isolation.

  • Python
  • FastMCP
  • ChromaDB
  • SQLite
Explore RAG-MCP

Desktop AI / agents

Gnome Agent

A local-first AI assistant inside the GNOME desktop. Combines multi-provider LLM support, MCP tools, streaming responses, and persistent conversations.

  • LangGraph
  • FastAPI
  • MCP
  • SQLite
Explore Gnome Agent
Implementation notes

A GNOME Shell extension connects to FastAPI over HTTP and server-sent events. LangGraph handles agent execution; SQLite checkpoints preserve conversation state across restarts. MCP connects desktop and filesystem tools.

Retrieval / conversational AI

RAG Chatbot

Chat with your own text and CSV data. A retrieval-augmented API that combines local Ollama models, Hugging Face embeddings, and PostgreSQL vector search.

  • LangChain
  • Ollama
  • pgvector
  • FastAPI
Explore RAG Chatbot
Implementation notes

Separate API endpoints initialize the vector store from text or CSV and handle chat queries. LangChain connects retrieval to an Ollama-hosted language model, with conversation history management.

Local inference / model serving

OpenVINO API Server

Local LLM inference through an OpenAI-compatible API. Uses Intel OpenVINO to serve chat completions with familiar request formats and streaming support.

  • OpenVINO
  • Python
  • FastAPI
  • OpenAI API
Explore OpenVINO Server
Implementation notes

Exposes /v1/chat/completions, /v1/models, and /health. Supports configurable sampling, token limits, and server-sent event streaming with locally stored OpenVINO model and tokenizer files.

AI-assisted infrastructure

Kubectl AI Agent

Translate natural language into Kubernetes commands. Separates command generation from optional execution, with validation, caching, rate limiting, and metrics.

  • Kubernetes
  • LangChain
  • FastAPI
  • Prometheus
Explore Kubectl AI Agent
Implementation notes

Distinct /kubectl-command and /execute endpoints keep generation and execution separate. The service includes API-key authentication, TTL caching, rate limiting, and Prometheus-compatible instrumentation.

Generative AI / image workflows

Stable Diffusion Studio

A Gradio interface for text-to-image generation and image styling. Connects Stable Diffusion with ControlNet, with controls for prompts, sampling steps, and denoising strength.

  • Stable Diffusion
  • ControlNet
  • Gradio
  • Python
Explore the repository

Applied AI / backend systems

Context-Aware Email Replies

Generate replies using email-thread context, stored summaries, and an LLM. Includes input and output guardrails, a raw-email webhook, and containerized Kubernetes deployment.

  • FastAPI
  • PostgreSQL
  • LLMs
  • Helm
Explore the repository

Project descriptions are based on the linked repositories.

Browse all repositories

02 / Technical focus

The model is only
part of the system.

My work connects AI application development with the data, deployment, and observability layers around it.

See it in the projects

AI applications & agents

Tool-using assistants, retrieval-augmented chat, persistent memory, and integrations with local and hosted language models.

Python / LangChain / LangGraph / MCP / Ollama

Data & model serving

Vector retrieval, document ingestion, API development, and local inference behind interfaces applications can use.

PostgreSQL / pgvector / ChromaDB / FastAPI / OpenVINO

Infrastructure & operations

Containerized workloads, Kubernetes deployment, infrastructure as code, CI/CD workflows, and service monitoring.

Docker / Kubernetes / Terraform / AWS EKS / Prometheus

03 / Professional background

Experience & education.

A systems-engineering foundation, with a focus on generative AI and practical applications.

Creative Synergies Group

Current role

Senior Software Engineer

Jul 2026 – Present

Bengaluru, Karnataka, India

OS3 Infotech Pvt. Ltd.

2022 – 2026

Engineer - Systems

Mar 2025 – Jul 2026

Mumbai, Maharashtra, India

Assistant Engineer - Systems

Jun 2023 – Mar 2025

Navi Mumbai, Maharashtra, India

Trainee Engineer - Systems

May 2022 – Jun 2023

Navi Mumbai, Maharashtra, India

Full professional profile on LinkedIn

04 / A little about me

Curious by nature.
Practical by design.

Bengaluru, Karnataka, India

I’m Ankit, an AI engineer with a DevOps background, based in Bengaluru. My focus spans generative AI, agentic workflows, retrieval-augmented generation, and MLOps.

I’m interested in connecting AI capabilities with practical applications: assistants that use tools, retrieval systems that bring in the right context, and software that fits into everyday workflows.

My background in systems engineering brings an infrastructure perspective to that work. Alongside AI development, I work with automation, containers, Kubernetes, deployment pipelines, and monitoring. My public projects show how those interests come together.

Practical AIThoughtful infrastructureOpen-source exploration

05 / Start a conversation

Have something
worth building?

Let’s talk about AI systems, engineering opportunities, or an open-source collaboration.

ankit.vish577@gmail.com
LinkedIn GitHub