AI Systems Engineer
Aaryan Patwardhan
I build systems that see, decide, and heal themselves.
About
Systems thinker.
Vision AI and autonomy.
B.E. Information Technology student at SPPU (2027). I design autonomous AI pipelines — from real-time computer vision at 55fps to self-healing drone fleets. I work at the intersection of deep learning, systems design, and edge inference.
Experience
What I've built
AI Core Lead
2026SentinelMesh — INSPIRON 5.0 (CSI COEP)
- —Designed MAPE-K autonomic loop for self-healing autonomous drone fleet
- —YOLOv8n at 55fps on RTX 3050 Ti with dual CUDA co-inference (vision + LLM)
- —Built Adversarial Debate Engine: Agent-A Dispatcher vs Agent-B Skeptic before any dispatch
- —Implemented confidence tier gating — LLM bypassed at conf ≥ 0.88 for ~50ms dispatch
AI/ML Engineer
2026Ghost-Admin — Autonomous Server Healing Agent
- —Designed MAPE-K feedback loop for autonomous server diagnosis and remediation
- —Integrated local LLM inference via llama-cpp-python with CUDA acceleration
Backend Engineer
2026Upwork Automation Pipeline
- —Built two-version automated job-hunting pipeline with LLM scoring and Telegram alerts
- —V2: composite scoring (50% AI relevance, 25% client quality, 25% competition opportunity)
Projects
Things I've shipped
SentinelMesh
Autonomous drone fleet that detects, debates, and dispatches — without human input.
Autonomous Upwork Pipeline
Automated job-hunting with LLM scoring, client quality signals, and Telegram delivery.
Ghost-Admin
Server healing agent that monitors, diagnoses, and self-repairs without human input.
PocketLawyer Edge AI
On-device legal assistant for Android with fully local LLM inference — no server, no data leaks.
PPE Detection System
Real-time safety compliance detection for industrial environments at 60fps.
Student Attendance Analytics
Automated attendance monitoring with analytics dashboard for educational institutions.
Full-Stack E-Commerce Platform
End-to-end online store with inventory management, cart, and order processing.
Skills
The stack
Hover the constellation nodes to explore skill co-occurrences
Vision AI
55fps real-time detection on edge hardware
Real-time frame pipeline with CUDA-accelerated preprocessing
Machine Learning
Custom training pipelines, mixed precision, CUDA streams
Local GGUF inference, Qwen2.5-1.5B Q4_K_M at production latency
Languages
Primary language across all ML and backend work
Backend
Full-stack Python/Flask with REST APIs and template rendering
Async WebSocket server for real-time drone fleet coordination
Leaflet.js dashboard with WebSocket real-time marker updates
Lightweight persistent dedup and feedback storage
Systems & Infra
Dual CUDA inference: YOLOv8 + GGUF LLM simultaneously on RTX 3050 Ti
Scripted automation and system tooling on Garuda Linux / Arch
Primary development environment; deep kernel and driver familiarity
Mobile
Edge AI app with local LLM inference on-device
Contact
Let's build something.
Available for freelance AI/ML engineering work. Response time: < 24 hours.