BHarath_07 FreshMindAI

Hi 👋, I'm Bharath Pemmanaboyana

💼 Software Engineer 2 at MAQ Software, Hyderabad, India — 3+ years building end-to-end ML & agentic AI systems.
⚙️ I design production-grade agentic architectures, build MCP servers, and orchestrate multi-agent workflows with LangGraph.
🧠 Strong foundation across both modern LLM systems (agentic pipelines, DSPy, vLLM) and classical ML/NLP (classification, ranking, retrieval, information extraction).
📊 Rigorous about evaluation — A/B testing, statistical hypothesis testing, and latency/cost/quality trade-off analysis in every system I ship.
🛠️ Comfortable across the full stack: Python, C#, C++, SQL, Docker, CI/CD, and Azure cloud infrastructure.

🤖 Autonomous SDLC Agent Framework — automating code generation, PR review, and DevOps pipelines end-to-end.
🌐 Multi-Agent Orchestration — stateful, graph-based workflows across real-time data sources using LangGraph and the Microsoft Agent Framework.
🔌 MCP Server Development — building and refining Model Context Protocol servers for Dataverse and enterprise data integration.
⚡ Local SLM Orchestration — fine-tuning and evaluating sub-31B models (e.g. Gemma) with DSPy and vLLM, optimizing for latency, cost, and reliability.

Generative AI & Agentic Systems

Machine Learning & NLP

Languages & Frameworks

Data, Cloud & DevOps

Aizen Trading — Multi-agent AI options trading system (LangGraph, XGBoost, GNN, Alpaca API). 8 specialized agents orchestrated behind a deterministic risk engine, with a provider-agnostic Anthropic/OpenAI LLM interface and an immutable decision journal for auditability.
Multi-Agent Orchestration with LangGraph & MCP — Stateful, graph-based agent system with agent memory, reflection loops, and self-correction, improving task completion accuracy by 30% over a single-pass LLM baseline.
Diabetes Prediction — End-to-End ML Pipeline — Full classical ML pipeline (EDA, feature engineering, model comparison, A/B-tested variants) evaluated via ROC-AUC/F1 prior to deployment.