I'm a Computer Science student specializing in Artificial Intelligence and Machine Learning at Jain University, Bengaluru.

My work generally sits between AI engineering and backend development: taking models or LLM workflows beyond notebooks and turning them into usable systems with APIs, persistence, retrieval, concurrency, testing, and deployment.

Currently interested in:

• Agentic AI and tool-using systems
• Retrieval-Augmented Generation and semantic search
• Speech and multilingual NLP systems
• Backend/API architecture with Python and Java
• Local and resource-efficient AI inference

Offline multilingual speech-to-speech translation for 8 Indian languages.

Microphone → Whisper ASR → Argos Translate → MMS-TTS → Speech

Built as a concurrent four-stage inference pipeline rather than a sequential model demo.

• Runs completely locally after model setup; no cloud inference APIs
• Uses faster-whisper, Argos Translate and Facebook MMS-TTS
• Separate worker threads for audio capture, ASR, translation and synthesis
• Overlapping audio windows prevent words from being lost at chunk boundaries
• Sentence buffering preserves translation context across ASR fragments
• Bounded queues and drop-oldest backpressure keep the pipeline responsive
• Automatic CUDA/CPU selection for TTS
• Includes automated tests and ASR benchmarking infrastructure

Conversational business-intelligence agent over live Monday.com operational data.

Built an AI agent that can reason over Deals and Work Orders boards and answer questions such as pipeline health, overdue work, sector performance and leadership-level KPIs.

User
  → FastAPI
  → LangGraph ReAct Agent
  → Business Intelligence Tools
  → Monday.com GraphQL API

• LangGraph ReAct agent with tool-based reasoning
• Queries live Monday.com data through its GraphQL API
• Cursor-paginated board ingestion
• Cross-board analysis across sales and operational data
• Dedicated tools for pipeline, work-order and leadership analysis
• Handles inconsistent dates, currencies, missing values and dynamic schemas
• FastAPI backend with a separately deployed web frontend

RAG-powered gaming assistant with local retrieval, web fallback and persistent conversations.

Rather than sending every question directly to an LLM, the assistant uses progressively more expensive information sources:

Question
   → ChromaDB semantic retrieval
   → Web search when local knowledge is insufficient
   → Cache new knowledge
   → LLM synthesis

• ChromaDB-backed semantic retrieval
• Automatic web-search fallback using Serper and DuckDuckGo
• Successful web results are indexed for future retrieval
• CrewAI-based agent orchestration
• FastAPI backend
• Per-game conversation persistence using SQLite
• Graceful fallback when LLM rate limits are reached
• Embeddable browser sidebar interface

Team-built rental fleet intelligence system for tracking equipment, telemetry and utilization.

Developed as a hackathon project around rental-fleet operations and decision support.

• Asset check-in/check-out lifecycle tracking
• Telemetry ingestion and deterministic simulation scenarios
• Rule-based anomaly detection and alert generation
• Utilization analysis from engine and idle hours
• Demand forecasting and asset-allocation recommendations
• FastAPI + SQLAlchemy backend
• React frontend connected to the real backend API
• Reproducible seeded demo environment with automated backend tests

Multi-agent email triage and inbox automation system.

Uses specialized agents to process an inbox as a workflow rather than treating email automation as a single LLM prompt.

• Email categorization and priority classification
• Gmail organization and labeling
• Draft response generation
• Slack alerts for high-priority messages
• Rule-based cleanup and preservation policies
• Multiple configurable LLM providers

Full-stack multi-room chat application with a React frontend and Java Spring Boot backend, using WebSockets for real-time messaging.

Languages

AI / ML

Backend

Data

Engineering

Frontend

Java Developer Intern — EazBytz Worked on Java/Spring Boot backend development and reusable application components.

Tech Co-Lead — Cognito Club Contributed to technical development and helped conduct an Agentic AI workshop using Lyzr AI.

I'm particularly interested in systems where AI is only one component of the architecture — retrieval, state, tools, APIs, concurrency, evaluation and failure handling matter just as much as the model itself.

I'm currently looking to deepen my work in AI engineering, backend systems, agent infrastructure, and production-oriented ML.