AI Engineer · Agent Architect · LLM Systems Builder
I build agentic AI systems that actually ship to production — not just notebooks and demos. Multi-agent pipelines, real-time voice agents, LLM orchestration, backend APIs. I connect the dots across NLP, automation, and backend — end to end.
I came to AI automation late — but when I got it, I went all in. Not long ago I was still figuring out what agentic even meant. Today I am shipping multi-agent pipelines, real-time voice agents, autonomous workflow systems, and full-stack LLM-powered products — all live, all solving real problems for real users.
What changed? I stopped just learning AI and started building with it. Every project I build is deployed, tested, and solves a real problem — not just a portfolio piece.
Currently interning as an AI Application Developer at HandiCrafts, building LangChain-powered healthcare chatbots with RAG pipelines. Previously interned at IIITDM Kurnool building voice-based payment systems.
"Build things that solve real problems. Deploy them. Iterate."
Faith meets engineering. Sends prayer reminders 15 min before each Salah via Telegram, follows up on compliance, and returns a context-aware hadith from Sahih Bukhari via Claude API if a prayer is missed.
Real-time voice agent for veterinary emergency triage and appointment booking. Uses WebRTC audio streaming, LLM reasoning via ReAct paradigm, and ElevenLabs TTS. Handles mid-conversation context switching.
5-node autonomous agent swarm that classifies, routes, and acts on emails. Auto-Unsubscribe module detects spam links and triggers browser automation. Adaptive feedback loops for continuous learning.
Production-deployed AI ordering system for a live restaurant. LLM-powered dish Q&A, neural voice descriptions, 4-tier architecture, and automated dual email dispatch to customer and kitchen on every order.
Fetches from arXiv, Semantic Scholar, HuggingFace Papers, and Google Scholar. AI-driven ranking and scoring, persistent deduplication memory, real-time feedback loop via Telegram buttons.
A semantic search and Q&A system built over 101,502 authenticated Islamic source chunks — Quran (3 translations), 4 Hadith collections, and Tafsir Ibn Kathir. Question classifier detects emotional vs factual queries. Every answer traced to authenticated sources with zero hallucination.
A 6-agent sequential pipeline (Goal Analyst, Tech Advisor, Planner, Coder, Tester, QA) that takes a plain English description and delivers a complete, ready-to-run project as a ZIP file. Runs entirely on zero-cost infrastructure. Solved Groq 6000 TPM rate limits through per-agent crew isolation and architectural redesign.
Multi-agent AI system for real-time financial transaction analysis, regulatory validation, and automated risk scoring. JWT-secured role-based dashboard with REST APIs and CSV/PDF reporting.
3-agent swarm analyzing GitHub activity, resume text, and coding profiles to compute weighted technical readiness scores. FastAPI orchestrator aggregating outputs into a unified placement dashboard.
Unified agentic framework combining n8n with 9 open-source HuggingFace models. RAG pipeline with vector stores, 60-70% ticket resolution reduction, 80-95% API cost reduction vs proprietary alternatives.
Open to full-time roles and internships in AI Engineering, Agentic Systems, and Backend Development. If you're building something meaningful — let's talk.