Ahmedabad, India
2027 // B.E. AI & ML

orchestrating agentic systems at scale.

Mastering state-of-the-art architectures: driving high-performance deep learning models, autonomous multi-agent orchestration, and production-grade backend security.

role AI/ML & Agentic Engineer
status exploring boundaries // always innovating
MASTERING
Agentic Systems Deep Learning Full-Stack AI Data Cryptography
explore
profile

Professional Registry // Bio

Orchestrating Complex Agentic Operations.

Operating at 100% capacity in Artificial Intelligence & Machine Learning at L.D. College of Engineering. I build secure, high-performance deep learning architectures and tool-calling agent systems, backed by nature-inspired optimization algorithms and strict cryptography rules.

Input
Planner
Tools
RAG
Groq LLM
SYSTEM CORE: ACTIVE
SYS-01 // AGENT-NET
> init agent_network_core...
> system handshake [secured]
> agentic matrix: online
chronicle

Academic Credentials & Foundations

2023 - PRESENT
L.D. College of Engineering
Specialization in AI & ML
B.E. Specialization in Artificial Intelligence & Machine Learning (6th Semester).

Driving deep research in Neural Networks (backpropagation, activation optimization, multi-layer perceptrons), nature-inspired swarm intelligence algorithms, socket networking paradigms, and cryptosystem implementations.

neural nets cryptography swarm logic sockets
missions

Active Hero Projects

LLMBench Eval

mission_01 // AI Quality Engineering
jyotiraditya21-bug.github.io/LLMBench

A production-grade AI quality engineering platform designed to benchmark Large Language Models, evaluate prompt templates, detect regression anomalies, and optimize LLM query costs. Powered by an asynchronous Celery and Redis pipeline with LLM-as-a-Judge evaluations and automated failure diagnostics.

Asynchronous Celery/Redis execution engine coupled with LLM-as-a-Judge evaluation metrics and automated prompt injection audits for continuous LLM regression testing.
llm evaluation celery redis fastapi postgresql prompt testing plotly.js

Luminate AI RAG

mission_02 // Advanced RAG
luminate-rag active

> parsing arXiv research literature...

> RAPTOR index: built hierarchical tree [ok]

> running CRAG relevance grader...

> grading nodes: ambiguous context [fallback active]

> corrective query arXiv API: fetch deepseek-v3...

> final synthesis: context resolved [100% correctness]

A client-side serverless search assistant designed to synthesize research papers. Resolves context fragmentation by building a RAPTOR hierarchical clustering index and leverages Corrective RAG (CRAG) to fall back on live arXiv literature when local context is insufficient.

RAPTOR hierarchical tree summarization coupled with Corrective RAG (CRAG) fallback mechanisms, resulting in 100% corrective search success.
raptor corrective rag arxiv api next.js typescript openai api

RepoMind Agent

mission_03 // Agentic AI
Jyotiraditya21-bug / RepoMind
● scanning
RepoMind public
├─ backend/
│ ├─ agents/
│ │ ├─ summarizer.py
│ │ ├─ architect.py
│ │ ├─ critic.py
│ │ └─ rag_indexer.py
│ ├─ main.py
│ └─ routes.py
├─ frontend/
│ ├─ index.html
│ └─ app.js
├─ requirements.txt
└─ README.md

An AI-powered codebase onboarding assistant that analyzes public GitHub repositories to generate interactive dependency graphs, architectural designs, and custom developer onboarding checklists — powered by a self-correcting multi-agent LangGraph pipeline.

5-agent LangGraph orchestration: Summarizer → Architect → Critic (self-correcting loop) → RAG Indexer → Q&A Router for zero-hallucination developer onboarding.
langgraph fastapi ast parsing rag groq cloud github api

MAC-SPOT CLI

mission_04 // AI Tooling
mac-spot-shell active

> mac-spot --explain "IndexError: list index out of range"

> analyzing stack trace and source context...

> fetch: Gemini API model response [ok]

> executing code check: no syntax errors

> interactive chat mode: initialized

A production-ready macOS command-line assistant built in Python and powered by Google's Gemini API. Provides developers with real-time CLI debugging, error explanations, code reviews, automated Git commit message generation, and conceptual coding cheat sheets.

Seamless macOS shell integration, Gemini-driven code analysis, and interactive developer CLI utilities.
gemini api python macos cli click rich

Lead-Bot AI

mission_05 // Autonomous Conversational Sales
lead-bot // core
9ms latency
Hi! I can help you hire Jyotiraditya or co-engineer a project. What are you building?
Hey! Looking for an Agentic AI dev to build a custom RAG agent.
Type message...

An autonomous conversational agent built for portfolios and SaaS interfaces to qualify leads, answer project/technical FAQs, and capture contact signals. Powered by a local rule-based intent router and vector matching algorithms for zero latency.

Multi-turn dialogue state management with intent routing (Hire / Project / FAQ) and automated lead capture notification via webhooks.
conversational ai dialogue management lead capture fastapi websockets emailjs
services

Autonomous Engagement // Conversion Matrix

Qualify Leads 24/7 Autonomous Agent.

Lead-Bot AI is a custom conversational assistant designed to capture, qualify, and convert portfolio visitors in real-time. It acts as an interactive sales layer—answering technical questions, showing credentials, and capturing project briefs.

89.4%
engagement lift
142
leads qualified
< 10ms
local routing
Context-Aware Intent Classification: Automatically routes users to Hire, Project Quote, or general QA paths.
🔒
Direct CRM/Email Transmission: Dispatches validated lead records instantly using secure client-side handlers.
AGENT CORE: ACTIVE
SYS-02 // LEAD-BOT
Hello! I am Jyotiraditya's autonomous Lead-Bot. I qualify inquiries and transmit signals directly to his inbox. How can I help you today?
>_ AGENT REASONING TRACE (TELEMETRY)
> init lead_bot_agent_network...
> dialog status: idle // waiting for user path
transmit

Get in touch // Start a conversation

Let's talk projects.

Have an active mission, deep learning research, or an autonomous agent pipeline you want to co-engineer? Transmit a signal, and let's go beyond limits.