AI Agent Engineer · Full-Stack Developer · 3D Web
I build autonomous agents that actually do the work — and the interfaces people trust them through.

Frontend developer turned AI builder. I spent my early career shipping UI for financial and enterprise web apps at Vanguard and Infosys, then went deep on agent systems — and now I build tools that plan, decide, and execute with a human in the loop.

• 🤖 AI agents & orchestration — MCP tool servers, multi-tool agents, approval gates, local + hosted models
• ⚙️ Automation workflows — n8n pipelines, LLM scoring, Gmail/Slack/Telegram as control surfaces
• 🧱 Full-stack — React, Node.js, TypeScript, REST APIs, OAuth
• 🌐 3D web — interactive Three.js experiences and motion-driven interfaces
• 🎓 M.S. in Artificial Intelligence (in progress) · M.S. in IT Management
• 📜 Anthropic Claude 101 · AI Fluency: Framework & Foundations
• 🧠 I learn by building — every project here started as something I actually needed

Languages

AI & Agents

Frontend & 3D

Backend & Infra

A personal AI chief-of-staff you operate entirely from Telegram — no chat UI, no dashboard.

Igris runs locally and orchestrates real work across my inbox, browser, and job pipeline: Gmail summaries, job scouting, resume tailoring, and application tracking — all exposed as MCP tools.

• Built as a Concierge Agent for the Kaggle × Google AI Agents Intensive capstone
• Human-in-the-loop by design — never applies to a job, uploads a resume, or sends/deletes email without explicit confirmation
• Structured into skills, specs, policies, tests, and evals so its behavior is testable and auditable
• Node.js · TypeScript · MCP · Telegram Bot API · Gmail API / OAuth 2.0 · Playwright · Ollama (gemma3:4b) · PM2

An open-source, local-first desktop + CLI workspace for AI-assisted coding — bring your own model, bring your own agent.

Most agent IDEs lock you to one vendor. AgentForge treats models, agents, and tools as pluggable through a shared Agent Protocol, with a permission engine that keeps risky actions gated behind you.

• Pillars: bring-your-own-model · bring-your-own-agent · local-first · human-in-the-loop permissions
• Monorepo: desktop app, CLI, and a shared agent core
• Planned stack: Tauri 2 · Rust · React · TypeScript · Monaco · xterm.js · SQLite

Universal data ingestion and persistent memory for AI agents.

Give an agent long-term memory over your own files and knowledge. Ships with a CLI, a knowledge-graph builder, and Claude Code skill integration.

• Python · Knowledge graphs · CLI · Claude Skills

A fully automated daily job pipeline built in n8n — scrape, score, tailor, deliver.

Runs on a morning trigger: pulls my resume, builds targeted search URLs, scrapes listings, ranks them with two-stage LLM scoring, generates an ATS-optimized resume per role, and delivers the digest by email with approvals routed to Telegram.

• n8n · Docker · LLM scoring · Scrapling · Google Drive/Docs API · Gmail · Telegram

An interactive 3D portfolio — motion, depth, and performance on the web.

• React · Three.js · TypeScript · Tailwind

• Building AgentForge Studio toward a first public release
• Extending Igris into goal-loop autonomy — give it an objective, it iterates until done
• M.S. in Artificial Intelligence — agent architectures, evaluation, and applied LLM systems
• Open-sourcing more agent tooling: MCP servers, Claude skills, and reusable automation blueprints

I'm open to roles and collaborations in AI engineering, agent systems, and applied LLM products.