My Work

Building Smart Automations with AI, Data & Workflow Intelligence.

My Projects

Interactive Multimodal AI Buddy – An AI Companion for Real-Time, Emotion-Aware Conversations

Architected an emotion-aware AI companion (Electron + FastAPI) with a dual-WebSocket pipeline bridging Gemini 2.5 flash audio streaming, a LangGraph reasoning pipeline (LoRA + DPO fine-tuned Mistral 7B), FaceNet-PyTorch facial auth, and pgvector semantic memory at ∼5s conversational latency.

Multi-Agent SystemsVision AnalysisAgent ReasoningLoRA Fine-TuningReinforcement Learning

AI Nutritional Health Assistant: Personalized Guidance for Indian Diets

Designed an agentic nutrition assistant (FastAPI + Next.js, LangGraph) with two-stage hybrid retrieval (BM25 + FAISS→RRF→cross-encoder re-ranking), 25+ health metric computations, and Gemma-3-27b via NVIDIA NIM- achieving 90% accuracy vs ∼62% naive search across 100+ real-world health scenarios.

FastAPILangGraphHybrid RAGFAISSLLM

Personal Finance System for Expense Tracking and Analysis

Engineered a containerized ML platform (FastAPI + Node.js, Docker Compose) with multivariate anomaly detection (LOF, One-Class SVM, PyTorch autoencoder), category-level forecasting (XGBoost/RF/GBM), Jensen Shannon drift detection, and semantic clustering via REST APIs- processing 10,000+ transactions in ∼3s.

PyTorchMachine Learning AlgorithmsStatistical Data AnalysisREST APIsDocker

Administrator Controlled Timetable Based Automatic Facial Recognition Attendance System

Developed an ESP32-CAM-based facial recognition system with OpenCV for automated, timetable-linked attendance of 50+ students, achieving 85% accuracy and winning Best Presentation at SYMPOSIUM-2023.

PythonComputer VisionPattern RecognitionESP32 CAMMySQL

Upcoming Projects

Home Data Server

Cloud Data Systems that can be hosted in home to provide various services like Media Streaming, File Storage, Automation etc.

UbuntuRaspberry PiData EngineeringDockerAutomation

Multi Modal AGI

Working on the MultiModal Assistant for Human like Experience!

UbuntuRaspberry PiIoTDockerAutomation