Independent Consulting
Ship your AI system —
not just a pilot.
I'm Girish Prasad Sahu — independent AI consultant with 17+ years building production-grade systems. I help companies design and deliver MCP agents, RAG pipelines, and LLM architecture that holds up at scale. Founder of Ajatus Software, scaled to 50 engineers. Currently consulting with Zeon AI; previously Lead Architect at TransUnion CIBIL.
Services
How I Work With You
Production AI Implementation
Full-cycle design and delivery of AI systems that run in production — MCP agents, RAG pipelines, LLM orchestration, and the infrastructure around them. Not a prototype. A system that holds up.
- Architecture design & technology selection
- MCP server development and integration
- RAG pipeline design with vector store selection
- Production deployment, monitoring, and handover docs
AI Strategy & Architecture Review
One-off deep-dive into your current AI architecture or roadmap. I identify the gaps, risks, and what to build next — based on 17 years of production systems experience and hands-on AI implementation.
- Current architecture audit with written findings
- Risk and gap analysis
- Prioritised build roadmap
- Model and infrastructure recommendations
Fractional AI CTO
Ongoing technical leadership for companies that need senior AI architecture guidance without a full-time hire. I work alongside your team, review decisions, unblock critical problems, and keep your AI direction coherent.
- Weekly architecture and strategy sessions
- Code and design reviews
- Vendor and model selection guidance
- Engineering team mentoring
Case Studies
Problems Solved
Giovani Textile, Thailand
Problem
Business team could not access operational data without developer help. Each query took 1–2 days of back-and-forth.
What I Built
Production MCP agent on Claude connected to live PostgreSQL. ChromaDB schema embeddings for natural-language query understanding. Zero developer involvement in the query path.
Result
Business team runs 20+ queries per day against production data in plain English.
TransUnion CIBIL
Problem
Client inquiry analysis and root cause investigation required manual Splunk and Solr log review — slow, error-prone.
What I Built
RAG pipeline connected to Splunk and Solr consoles. LLM-orchestrated root cause analysis with MoM variance tracking.
Result
Analysis time reduced from hours to minutes. Deployed under strict RBI compliance.
E-Challan (State Government)
Problem
Manual traffic violation review across hundreds of camera feeds — impractical at scale.
What I Built
Custom YOLOv5 model trained from scratch. FastAPI inference APIs with RabbitMQ image queue. Integrated with Parivahan for automated challan generation.
Result
Real-time violation detection pipeline. Automated challan generation end-to-end.
FAQ
Common Questions
Do you work with early-stage startups or only enterprises?
Both. I have worked with solo founders building their first AI product and enterprise teams at Flipkart, IndiGo, and TransUnion CIBIL. The engagement structure adapts to your stage.
What does a typical engagement look like?
Most project engagements run 4–8 weeks. We start with a scoping call, agree on deliverables, and I work in focused sprints. You get working software, architecture documentation, and a handover — not slide decks.
Can you work with our existing engineering team?
Yes. Most engagements involve working alongside your developers — reviewing code, unblocking architecture decisions, and transferring knowledge so your team owns the system after I leave.
What tech stacks do you work with?
Python, Node.js, Java, TypeScript on the backend. PostgreSQL, ChromaDB, MongoDB for data. Anthropic Claude, OpenAI, Ollama, Gemini for AI. AWS for infrastructure. I pick the right tool for the job, not the familiar one.
Get in Touch
Ready to start?
Describe your project — what you're building, where you're stuck, or what you want to achieve. I reply within 24 hours.
Connect
linkedin.com/in/girishprsahu