AI Full-Stack Engineer
for Hire
Generative AI & LLMs · RAG Pipelines · Agentic Workflows · FastAPI
I'm Ramesh Kumar Das. Over 6.5+ years I've built AI features that sit inside real products — backends, web apps, and the glue between them. My focus is Generative AI, LLMs, RAG pipelines, and Agentic AI workflows: the parts that have to work under load, not just in a notebook. At Kliky AI (Kuwait) I lead AI on WinstaAI, our live agentic SaaS with multi-model routing, text-to-image, video AI, and autonomous jobs. If you need someone who can go from prompt design to a deployed API, let's talk.
AI & Full-Stack Tech Stack
Tools I use from model wiring through deploy
AI Engineering Services
Concrete work I take on when a product needs AI built in
🤖 Generative AI Platform Development
Full AI SaaS builds: LLM hooks, credits and billing, admin panels, usage stats, and routing across several models. FastAPI services on the back, Next.js on the front.
🔍 RAG Pipeline Architecture
Retrieval stacks with LangChain or LlamaIndex, vector stores like Pinecone or Weaviate, and semantic search so the model answers from your own docs — not generic training data.
🧠 Agentic AI Workflows
Agents that call tools, hit APIs, and run multi-step jobs with light human oversight. LangChain agents, tool calling, and custom orchestration where frameworks fall short.
🎨 AI Media Generation Systems
Text-to-image (ComfyUI, Replicate, FAL.ai), face swap, video and audio processing, OCR, and speech-to-text / TTS. GPU runs on RunPod and similar hosts.
🌐 AI-Powered SaaS Backends
APIs built for AI workloads: Celery queues, WebSockets for streaming output, usage meters, rate limits, and sensible GPU budgeting.
🔗 AI API Integration
Wire OpenAI, DeepSeek, Anthropic, Hugging Face, and others into apps you already have. Prompts, response checks, cost control, and fallbacks when a provider flakes.
Work Experience
Where I've shipped AI into live systems over 6.5+ years
WinstaAI — Agentic AI SaaS Platform (Kuwait)
I own the AI platform build: generative features, agent workflows, and routing across multiple LLMs in a product customers use every day.
- Shipped agentic AI pipelines for tool calling, branching decisions, and multi-step jobs with LangChain plus custom orchestration
- Connected text-to-image (ComfyUI, Replicate), face swap, OCR, video/audio AI, and model routing across OpenAI, DeepSeek, and Ollama
- Built the admin dashboard for service toggles, credits, usage stats, and billing
- Hooked up GPU inference on RunPod and FAL.ai with dynamic allocation so spend stays under control
Dibbly — Multilingual AI Video Translation | Paperport — AI Content Generator
- Built multilingual video translation with speech-to-text, TTS, DeepL, and live transcription
- Added LLM content generation that turns rough text into presentations and roadmaps
- Hardened API traffic with sessions, CSRF protection, and JWT auth around the AI endpoints
Khantailor & PathHub
- Shipped full-stack platforms with React.js, Node.js, and Django
- Added real-time sync with WebSockets and Redis (about 30% better synchronization in practice)
- Cut API latency roughly 20–30% with backend tuning and caching
Frequently Asked Questions
Questions I get from hiring managers and founders
Get in Touch
AI engineering roles, product contracts, and consulting — remote worldwide