Manindra Reddy

Linkedin

Manindra Reddy

Bapatla Engineering College

SALESRAGER (AGENT )

I’ve built an AI agent that combines traditional machine learning models with personal business data, integrating them through LangChain tool functions and using a Mistral LLM (via Ollama) as the reasoning engine. This architecture gives my project a unique edge. The ML component consists of two parts, each serving a distinct purpose: Sales Forecasting – A Facebook Prophet model predicts revenue for specific time windows such as 1 month, 4 months, or 6 months. Lead Conversion Prediction – An XGBoost classification model predicts the likelihood of a lead converting. A common question is how inputs flow through the system and how the final outputs are generated. Here’s how it works: When a user provides input—such as selecting a forecasting period (e.g., “next one month”)—the LangChain tool function structures this query into a validated JSON schema. This schema is then passed to the Mistral LLM, which interprets the request and triggers the correct backend ML model (Prophet or XGBoost). The sophistication of the system comes from how reliably LangChain maintains schema validation and ensures the LLM always sends accurate, structured inputs to the machine learning models. This creates a seamless interaction between natural language queries and precise model-driven outputs. If you'd like, I can also rewrite t

Yes

No

Yes

Yes

Yes, I’m interested