Krishna Singh
AI Systems Engineer
Production AI • LLM Engineering • Intelligent Systems • Visual Computing • Automating Workflows
Systems thinking, shipped to production.
I architect and ship production-grade AI systems — from retrieval pipelines and reasoning engines to autonomous agents and streaming backends. I care about the boring parts that make AI reliable: latency budgets, evaluation, observability, and clean system design.
Architect
Design AI systems for latency, cost, and reliability — not demos.
Build
Ship retrieval, reasoning, and agent runtimes on FastAPI backends.
Operate
Instrument, evaluate, and iterate on live production traffic.
An engineering timeline.
Engineering production-ready generative AI systems, agentic workflows, and retrieval-backed applications that turn complex business processes into reliable software.
- Designed and shipped LLM-powered applications with structured tool use, retrieval, and evaluation workflows.
- Built scalable Python and FastAPI services for agent orchestration, streaming responses, and production integrations.
- Improved reliability through observability, prompt and model evaluation, and practical guardrails for enterprise AI.
System Flow
QueryEmbedRetrieveRe-rankGenerateGenerative AILLMsAI AgentsRAGPythonFastAPI
Skills as a connected blueprint.
Hover a node to trace connected technologies.
Systems built to ship.
Gemini RAG Engine
A grounded question-answering engine with source-cited responses, built on Google’s managed File Search and Gemini 2.5 Flash.
AI Conversational Chatbot
A multi-turn conversational agent with intent routing, tool use, and real-time voice via streaming backend.
MCP CRM Chatbot
An MCP-powered conversational CRM assistant that turns natural-language requests into safe, actionable customer and pipeline workflows.
AI Interview Platform
An intelligent interview platform that combines structured candidate sessions, AI-generated questions, and automated feedback.
Face Recognition System
A real-time face detection and recognition pipeline with embedding-based matching and an inference API.
Architectures, animated.
Production AI Pipeline
Request enters an API gateway, routes through validation and rate limiting, hits the inference service with caching and observability, and returns a streamed response.
RAG Architecture
Documents are chunked, embedded, and indexed. A query is embedded, retrieved, re-ranked, and passed to the LLM for grounded generation.
Financial AI Platform
Real-time market and portfolio streams feed a pipeline that enriches data, runs AI reasoning, and persists insights to PostgreSQL.
Streaming AI Backend
A client connects over WebSocket, the agent runtime calls tools and models, and tokens stream back continuously with memory in Redis.
MCP CRM Agent
A conversational CRM request is interpreted by an agent, validated through MCP contracts, executed with scoped tools, and returned with an auditable result.
AI Interview Platform
Candidate and role context drive an interview workflow that generates questions, captures answers, evaluates competencies, and produces structured feedback.
The full technology stack.
All Technologies
Certifications.
Developer activity.
@krishnasingh-28
Live from GitHub
Let's build the next generation of AI systems.
Send a transmission
Designed & engineered by Krishna Singh · 2026