Munna Suprathik

AI Engineer building production agentic systems with memory, tool execution, semantic retrieval, and workflow automation. Developed conversational assistants, multi-agent RAG, SQL agents, and platform APIs using Python, Node.js, LangGraph, OpenAI, Docker, and OCI. Experienced in evaluation, automated testing, monitoring, and production deployment.

AI Engineer, R&D - Conneqtion Group

(2025-08)

• Built and deployed a production HRMS conversational assistant on Microsoft Teams and OCI for 400+ employees and 3,000+ monthly interactions, reducing HR workload by 45 to 60 hours per month.
• Reduced agent latency by 43%, from 7 seconds to 4 seconds, and cut errors by 70% through prompt tuning, retrieval optimisation, endpoint profiling, and Grafana monitoring.
• Built a multimodal semantic retrieval system with 78 to 82% top-1 accuracy across 500 to 700 documents and 1,500+ images using Gemini, Pinecone, and OCI.
• Developed a multi-agent RAG system over 48,000+ structured and unstructured records, using routed retrieval and tool execution to improve lead qualification by 20 to 30% and save 10 to 15 hours per week.
• Built a natural-language SQL agent with governed tool execution for purchase requisitions, orders, invoices, and receipts, giving users verified answers without writing SQL.

AI Engineer Intern - Conneqtion Group

(2025-02 - 2025-08)

• Built conversational workflows with live SQL queries, document retrieval, function calling, and REST API integrations, achieving a 6-second end-to-end response time.
• Validated 1,000+ invoices with 87% field accuracy and 90% stamp-detection accuracy using OpenCV contour detection and an AI scoring layer.
• Developed an evaluation harness for four foundation models across multimodal, RAG, and reasoning benchmarks, which became the internal standard for model selection.

B.E. - Computer Science and Engineering - New Horizon College of Engineering (2025-05)