An AI & Backend Systems Engineer specializing in the design and orchestration of multi-agent LLM workflows. The core engineering focus is on replacing single-threaded, blocking LLM calls with asynchronous, queue-based architectures that stay responsive under long-running inference.

Expertise lies in system-level orchestration using LangGraph and Google ADK to coordinate specialized AI agents, and structured outputs (Pydantic) to keep multi-step reasoning reliable rather than freeform. Rather than treating AI as a simple chatbot interface, the focus is on using AI as a reasoning layer embedded within decoupled services agents that plan, retrieve, and act across separate frontend, backend, and worker processes.

This technical execution includes deploying agent workflows over asynchronous message queues (Redis/BullMQ), maintaining persistent semantic memory through vector databases (Supabase pgvector, Pinecone), and containerizing backend and worker processes (Docker) for deployment across Render, Vercel, and Hugging Face Spaces.

Stack: Next.js 16 · FastAPI · LangGraph · Groq (LLaMA 3.3 70B) · Upstash Redis · Supabase PostgreSQL Deploy: Vercel (frontend) · Hugging Face Spaces Docker (backend + worker on port 7860)

• Distributed research engine that decomposes complex queries into parallel execution threads via a LangGraph agent swarm (Architect → Scouts → Synthesizer).
• Backend worker runs as a background process in the same Docker container, consuming jobs from Upstash Redis via BRPOP and archiving synthesized reports to Supabase.
• Scouts execute concurrent web searches via duckduckgo-search + Python asyncio, eliminating I/O bottlenecks.

Stack: React 18/Vite · Node.js/Express · Python/FastAPI · BullMQ · Upstash Redis · MongoDB Atlas · Firebase Auth Deploy: Single Docker container on Render free tier (Node + Python via concurrently)

• Full-stack career copilot — resume parsing (PDF → GridFS → LLM), job matching, cover letter generation, and interview prep, all offloaded to BullMQ workers.
• Python is a pure stateless HTTP service; Node BullMQ workers consume queue jobs and call Python over localhost HTTP. Python never touches Redis.
• Switchable LLM providers (OpenAI / Groq / Gemini) via AI_PROVIDER env var with 55s timeouts.

Stack: Next.js 16 · Python/FastAPI · Google Gemini (gemini-3.6-flash) · Pinecone Deploy: Vercel (frontend) · Render (backend: my-ai-brain)

• Multimodal AI engine processing text + image inputs through Google Gemini for inference, with persistent vector memory via Pinecone (vision-memory index) for long-duration context retention.
• Conversation state serialized to chat_history.json for local backup alongside Pinecone's durable vector layer.

Stack: Next.js 15 · TypeScript · shadcn/ui · Zustand · TanStack Query · Custom RAG pipeline Status: Local development prototype (not deployed)

• Local-first RAG engine for processing corporate RFPs — complete document ingestion pipeline with semantic chunking, embedding, and retrieval services.
• RAG service layer is fully implemented (7 TypeScript modules) but currently operates on mock embeddings; OpenAI integration is stubbed but not yet wired to a live key.

Stack: Next.js 16 · Supabase PostgreSQL · Google Gemini · NextAuth (GitHub OAuth) · n8n Status: Local development (not deployed)

• AI-powered code evaluation framework with a leaderboard dashboard. Google Gemini scores LLM-generated code against engineering rubrics.
• n8n webhook integration upserts evaluation metrics into Supabase via stored procedures, automating the scoring pipeline.

Stack: Google ADK 2.6 · Gemini 2.5 Flash · MCP Filesystem Server · Python Status: Local CLI tool

• Autonomous, self-healing file-system organizer that uses the Model Context Protocol (MCP) to restructure local directories via natural language. Engineered with a strict human-in-the-loop approval boundary (SCAN → PLAN → CONFIRM).
• Features Enterprise-Grade Graceful Degradation (a zero-downtime offline heuristic engine that instantly takes over if cloud LLM APIs fail) and a Global Undo Architecture for instant reversions of file migrations.

• GitHub Copilot Certified Architect (GH-300) — Microsoft / GitHub.
• IBM RAG and Agentic AI Professional Certificate — Coursera.
• Oracle Cloud Infrastructure Certified AI Foundations Associate — Oracle.
• Oracle Agentic AI Certified Foundations Associate — Oracle.
• Oracle Fusion AI Agent Studio Certified Foundations Associate (Rel 1) — Oracle.
• AI Engineer for Data Scientists Associate — DataCamp.
• Associate Data Engineer — DataCamp.
• Microsoft SQL Server Professional Certificate — Coursera.
• Google Advanced Data Analytics Professional Certificate — Coursera.

• IBM SkillsBuild & Edunet Foundation | AI & Cloud Engineering Intern

  • Participated in an intensive, project-based enterprise engineering track focused on emerging cloud platforms and agentic AI architectures under the IBM SkillsBuild initiative.
  • Designed and engineered the watsonx-nutrition-agent utilizing IBM Watsonx Orchestrate, implementing structured prompt behavior mapping, runtime knowledge-source ingestion, and automated multi-turn reasoning pipelines tailored for precision advisory domains.
  • Explored real-world enterprise agent integration patterns, leveraging cloud-native tools to orchestrate deterministic workflows and minimize hallucination vectors in conversational interfaces.

• NexLevr | AI & Automation Engineering Intern

  • Engaged in advanced automation engineering workflows, focusing heavily on execution resilience, system integration, and large language model output governance.
  • Developed and open-sourced the LLM-Output-Variance-CLI, an auxiliary developer tool engineered to systematically audit temperature parameters, measure probabilistic generation drift, and evaluate token-level variance across high-parameter inference endpoints.
  • Focused on bridging production reliability gaps by establishing reproducible measurement scripts for runtime model stability.

• System Design: Microservices, Event-Driven Queues, Container Orchestration (Docker), Serverless Functions.
• Data Infrastructure: Relational SQL, MongoDB (NoSQL), Supabase (pgvector), Redis (BullMQ), Vector Search (Pinecone).
• Backend Frameworks: Node.js, Next.js (Edge), Python (FastAPI).
• AI Integration: LLM Orchestration (LangGraph, Google ADK), RAG Architecture, Multi-Agent Swarms, Google Gemini, Groq.