SIDDHARTH DIVYANG PATNI
Building production AI systems that reason, retrieve knowledge, and control robots.
Agentic AI Engineer with production experience shipping autonomous scraping systems, LLM observability proxies, and multi-robot coordination platforms. Currently pursuing M.Sc. Digital Technologies at TU Clausthal, Germany.
ABOUT // EXECUTIVE SUMMARY
> Who I am, what I build, and what problems I solve.
EXECUTIVE_SUMMARY
I build production-grade AI systems spanning agentic workflows, robotics software, retrieval-augmented generation, and autonomous decision-making. My work combines software engineering, machine learning, and robotics to solve real-world automation problems — with experience across both research and industry.
Currently pursuing M.Sc. Digital Technologies at TU Clausthal & Ostfalia University, Germany. Previously shipped production systems at Ciconia Systems GmbH (CORE Research Group) and L&T Technology Services.
AGENTIC AI
Cascade pipelines, Computer Use Agents, self-learning route systems, and production LLM infrastructure.
SYSTEMS ENGINEERING
Distributed task queues, observability proxies, EU AI Act compliance, and MLOps pipelines.
ROBOTICS
Multi-robot coordination (Pepper + Temi), SLAM navigation, and IoT sensor networks.
EXPERIENCE // ENGINEERING TIMELINE
> Production systems shipped, research conducted, and teams contributed to.
AI Software Engineer (Research & Dev)
- Designed and implemented Vergabepilot.AI, an autonomous agentic AI platform to scrape and download public-procurement documents globally.
- Developed a 7-strategy cascade pipeline utilizing Playwright, LLM code-generation, and visual Computer Use Agents (browser-use).
- Created a self-learning route system that records agent clicks and replays them, reducing LLM costs by over 95% on repeat visits.
- Architected a fanned-out chunk processing system using Celery, Redis, and asyncio to process up to 10,000 URLs per job run.
- Configured Prometheus metric exports and Grafana dashboards for real-time scraper performance and cost tracking.
Software Developer Intern
- Refactored React components and Flask API layers, reducing perceived page load latencies by ~30% through bundle tuning and response caches.
- Dockerized microservices and integrated them into a Linux-based CI/CD pipeline, reducing environment-specific integration issues.
- Standardized API-UI interface contracts and added schema validation checks to prevent runtime errors.
Data Analyst Intern
- Analyzed property maintenance logs and spatial data arrays to support core product performance reviews.
- Generated SQL queries and automated reports to isolate task workflow delays and bottlenecks.
- Built visual reporting sheets and spreadsheets to track team ticket completion metrics.
Open Source AI & Systems Creator
- Engineered Spidey, an autonomous agent loop with custom React Flow diagrams, offline voice assistant (Vosk), and SFT/DPO fine-tuning.
- Built Sentinel, a security proxy providing EU AI Act audit ledgers (SHA-256 chained), span-tree Gantt charts, and model routing.
- Designed Cereforge, an AI/ML practice platform featuring LLM task evaluation, progressive hints, and an interactive AI mentor.
- Created Workflow Engine, a low-code distributed Bull queue task orchestrator with sandboxed Node VM and Python runtimes.
Robotics & Embedded Systems Developer
- Co-developed ShopMate-R, connecting Pepper and Temi robots to a central Flask orchestrator for grocery routing and checkout.
- Implemented a real-time weight sensor network using M5Stack IoT microcontrollers transmitting to Flask backend APIs.
- Built AI Driver Safety, integrating a traffic sign classifier (CNN), eye/drowsiness tracking (dlib), and speed tracking (OBD-II).
CASE STUDIES // ENGINEERING WORK
> Deep dives into architecture decisions, trade-offs, and production challenges.
Vergabepilot.AI
Autonomous agentic AI for public procurement document extraction across 30+ portal families globally.
CHALLENGE: Public procurement portals are highly fragmented, dynamic, and actively hostile to scraping (Cloudflare, captchas). Stan…
ShopMate-R
Multi-robot grocery shopping assistant using Pepper & Temi robots.
CHALLENGE: Navigating a complex grocery environment requires distinct physical capabilities: human interaction (Pepper) and payload…
Sentinel
Open-source LLM observability and agent reliability proxy with Cost & EU AI Act auditing.
Spidey
Self-hostable autonomous AI agent with voice controls, live reasoning graphs, and offline capabilities.
Cereforge
Interactive learning and practice platform for AI & MLOps engineers.
ENGINEERING STACK // CAPABILITIES
> Technologies used in production across agentic AI, robotics, and full-stack systems.
AGENTIC AI & AUTOMATION
PRIMARYAI, ML & COMPUTER VISION
PRIMARYFULL-STACK DEVELOPMENT
PRIMARYDATA & INFRASTRUCTURE
SUPPORTINGROBOTICS & HARDWARE
SUPPORTINGCERTIFICATIONS // ACHIEVEMENTS
> Professional certifications, research achievements, and notable coursework.
Fundamentals of Robotics & Industrial Automation
CORE Research Internship — Agentic AI
Multi-Robot Grocery Assistant — Pepper & Temi
Autonomous AI Agents & Multi-Agent Systems
Computer Vision & Deep Learning
Real-Time Operating Systems & Embedded IoT
High Distinction — Diploma in Computer Engineering
ACADEMIC RECORDS // EDUCATION
> Decrypting study profiles, research coursework, and academic credentials.
M.Sc. Digital Technologies
B.Tech. Computer Engineering
Diploma in Computer Engineering
CONTACT // SEND MESSAGE
> Open a channel for communication, collaboration, or project inquiries.
CONNECTION_PARAMETERS
Available for research initiatives, agentic frameworks, robotics, full-stack platforms, or collaboration opportunities. Let's connect.