I am an AI and Machine Learning professional focused on designing scalable, production-grade systems. My expertise lies in orchestrating multi-agent architectures, optimizing retrieval-augmented generation (RAG) pipelines, and implementing interpretable deep learning models. I specialize in bridging the gap between theoretical research and operational software, delivering robust solutions for complex technical challenges.

Real-Time AI Pedagogical Assistant

• Core Technology: React 19, FastAPI, LangGraph, Deepgram
• Description: An interactive teaching tool implementing the Feynman Technique through Socratic dialogue.
• Key Features:
  • Orchestrates complex conversational states using finite state machines to challenge user understanding rather than providing direct answers.
  • Features a decoupled architecture with a WebSocket-enabled low-latency audio pipeline.
  • Implements strict Voice Activity Detection (VAD) for natural, interruption-free turn-taking.

Autonomous Educational Video Generation Pipeline

• Core Technology: Python, Multi-Agent Systems, Manim, RAG
• Description: An end-to-end system that autonomously generates polished mathematical explainer videos from text prompts.
• Key Features:
  • Coordinates specialized agents (Solver, Evaluator, Developer) to ensure mathematical rigor and code correctness.
  • Utilizes a self-correcting "Golden Set" RAG mechanism to retrieve and validate animation code patterns.
  • Integrates VibeVoice for natural neural text-to-speech synthesis synchronized with generated visuals.

LLM-Driven Algorithmic Trading Executor

• Core Technology: Python, MetaTrader 5, OpenAI GPT-4o, Telethon
• Description: Automated trading engine that parses unstructured signals from Telegram channels to execute trades on MetaTrader 5.
• Key Features:
  • Uses GPT-4o to extract structured trade parameters (Symbol, Side, SL/TP) from natural language messages.
  • Implements strict risk management validation and stale-message filtering.
  • Maintains concurrent state for position management and duplicate signal prevention.

Semantic Code Search Engine

• Core Technology: Python, Tree-Sitter, FAISS, OpenAI Embeddings
• Description: A production-ready indexing tool for semantic search across large repositories.
• Key Features:
  • Implements AST-based chunking via Tree-Sitter to preserve semantic context of functions and classes.
  • Uses Merkle tree data structures to enable efficient, incremental updates of the vector index.
  • Supports over 25 languages with intelligent fallback strategies for non-code assets.

Autonomous Repository Analysis Agent

• Core Technology: Python, Google Gemini, ReAct Pattern
• Description: An intelligent agent capable of exploring and analyzing GitHub repositories to answer complex structural questions.
• Key Features:
  • Follows a Reasoning-Acting (ReAct) loop to autonomously plan and execute investigation steps.
  • Retrieves real-time file content and directory structures to ground answers in actual codebase facts.
  • Eliminates hallucination by relying on direct tool usage for information gathering.

Explainable AI (XAI) Investment Platform

• Core Technology: TensorFlow (LSTM), SHAP, Streamlit, Yahoo Finance
• Description: A comprehensive backtesting and trading analysis platform integrating Explainable AI.
• Key Features:
  • Deploys Long Short-Term Memory (LSTM) networks for predictive market modeling.
  • Integrates SHAP (Shapley Additive Explanations) to provide feature-level transparency into model decision-making.
  • Provides professional-grade metrics including Sharpe ratio, max drawdown, and profit attribution.

Clinical Risk Assessment Model

• Core Technology: Scikit-learn, Pandas, Streamlit, Tabula
• Description: A diagnostic support tool for predicting diabetes risk based on medical biomarkers.
• Key Features:
  • Implements Multiple Linear Regression trained on the PIMA Indian dataset.
  • Features a dual-mode interface for both manual data entry and automated PDF medical report parsing.

Languages

Frameworks & Libraries

Cloud & Infrastructure