Skip to content
View saiharsha-k's full-sized avatar
:octocat:
Building Production First AI Systems
:octocat:
Building Production First AI Systems
7 followers · 29 following

Achievements

Achievement: Pair ExtraordinaireAchievement: YOLOAchievement: Pull Sharkx2Achievement: Quickdraw

Achievements

Achievement: Pair ExtraordinaireAchievement: YOLOAchievement: Pull Sharkx2Achievement: Quickdraw

Highlights

Block or report saiharsha-k

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.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. 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
saiharsha-k/README.md

Sai Harsha Kondaveeti

AI Engineer · Production AI Systems · RAG · MLOps · AI Agents

AI is marketed as intelligent. It's deployed as fragile.

I am an AI/ML Engineer based in London. In production: shipped hybrid RAG pipelines, 15+ automation workflows orchestrating 30+ services, OCR automation processing 2,000+ resumes, and a lead scoring model contributing to 12% higher conversion at Datavalley Inc.

That obsession led me to work across the full production stack: classical ML pipelines, LLMs, RAG systems, AI agents, backend APIs, and cloud deployment — not as separate experiments, but as one connected system that either holds under pressure or doesn't.

Along the way I noticed two problems nobody was solving cleanly. When agents start making decisions on behalf of organisations, who is accountable? And when multiple agents work together, what stops them from doing too much — or too little?

That's what Agiorcx and Cordax are about.

Current Projects

Orion — Production AI Agency Platform

A chat-first, mobile-first PWA personal AI agency built on OPAF (Orion Production Agency Framework) — a custom 7-layer architecture solving five documented production failures in agentic systems.

Core innovation: a Reliability Layer with Silent Failure Detector, typed AgentResult contracts, and checkpoint-based rollback. Built with Python, FastAPI, Gemini 2.0 Flash, Qdrant, Supabase, MCP connectors, and real-time SSE streaming. Status: Active build · Private repo · Demo available on request

Agiorcx

The governance layer for the Internet of Agents — where autonomous systems coordinate with accountability, not just capability.
Deterministic coordination. Immutable contracts. Zero silent failures.

Cordax

The coordination brain for multi-agent systems — routing intent so each agent stays focused, testable, and accountable.
Cordax separates the decision of who acts from the act itself. Multi-agent systems fail because agents try to do too much. Cordax fixes that.

What I Build

  • End-to-end ML pipelines with MLOps practices
  • LLM applications and RAG pipelines for domain-specific systems
  • Prompt engineering frameworks: versioning, regression testing, evaluation
  • AI agent systems, orchestration patterns, and coordination design
  • Backend AI APIs with FastAPI
  • Cloud deployment on GCP (Vertex AI · Cloud Run) and AWS
  • Vector database and retrieval systems (FAISS · Pinecone · Weaviate)
  • Production trade-offs: reliability, latency, cost, monitoring, failure modes

Technical Stack

Languages & Core: Python · SQL
ML/DL: PyTorch · TensorFlow · Scikit-learn · Deep Learning · Model Evaluation
Generative AI: LLMs · RAG · Prompt Engineering · AI Agents · Fine-tuning
Frameworks: LangChain · LlamaIndex · Hugging Face · FastAPI
Vector DBs: FAISS · Pinecone · Weaviate
Cloud & Deployment: Google Cloud Platform · AWS · Vertex AI · Cloud Run · Docker
MLOps/LLMOps: Monitoring · Observability · Evaluation · CI/CD
Prototyping: Streamlit · Gradio · Jupyter

Education

MSc Data Science (AI & Machine Learning)
University of Roehampton, London

Writing & Publishing

Everything I build is documented publicly. Every repo has a matching post, article, or deep dive.

Open To

AI Engineer · Machine Learning Engineer · Generative AI Engineer · LLMOps Engineer · Data Scientist
London · Eligible to work full-time in the UK · No sponsorship required

Build reliable systems. Ship with intent.

Pinned Loading

  1. rag-axis rag-axis Public

    Opinionated, production-minded core library for RAG systems.

    Python 1

  2. production-ai-systems-engineering production-ai-systems-engineering Public

    Hands-on production AI engineering book — in active development. Covers RAG, agentic systems, MLOps, LLM engineering, and cloud deployment. Architecture-first, practitioner-written.

    1