AI-Based Predictive Maintenance and Operational Efficiency in Manufacturing
In preparation · targeting IEEE
Differentiation: Explainable AI (XAI) combined with lightweight edge deployment.
IEEEXAIEdge AIManufacturing
Berlin · MSc AI / Robotics
Saurabh Kumar
What if machines could feel?
Exploring the intersection of Robotics, AI, Quantum Computing, and Human Emotion. Building machines that understand feelings.
About
I'm an electrical engineer turned AI researcher and builder, based in Berlin. After four years shipping production software — leading engineering as Technical Head at Pashi Technology and launching Vocaa to users in 177 countries — I moved into research full-time. My aim is a PhD in Affective and Embodied robotics.
- Technical Head & 20% Partner — Pashi Technology (Led the engineering team)
- Full-Stack Developer — PGTech
- Data Analyst — Earlier role
- 4+
- years of full-stack engineering
- 177
- countries reached by Vocaa
- 20+
- talks on technology & AI
- EEE → AI
- hardware roots, software reach
Research
Research & Academic Work
Core Research Vision
What if machines could feel?
I believe the next frontier in robotics isn't just intelligence — it's emotion. My work explores Affective Robotics: building machines capable of empathy, emotional awareness, and genuine human-like connection. Drawing from Electrical Engineering, AI/ML, Quantum Computing, and Cognitive Science, I aim to pioneer systems where robots don't just respond — they understand.
Publications in preparation
Is Time an Illusion? Exploring the Emergent Nature of Temporal Experience
In preparation · targeting IEEE
A philosophical-computational paper arguing that time may be emergent rather than fundamental.
IEEEPhilosophy of PhysicsComputational Theory
Research interests
Affective Robotics
Giving robots emotional intelligence and empathy.
Quantum Robotics
Applying quantum computing to robotic decision-making.
Embodied AI
AI that lives in physical bodies and interacts with the real world.
Space Autonomy
Autonomous robotic systems for space exploration.
Open to PhD
Currently seeking PhD positions and research collaborations in Affective Robotics, Embodied AI, Quantum Computing, and Space Systems.
Projects
Built, not just proposed
Research-grade engineering in robotics and agentic AI, alongside production apps used by real people.
Robotics & AI Research Projects
FollowCam — Autonomous Camera Robot
Robot name: Kiko
50% Built — Presenting to NVIDIA, TU Munich (Oct 2026)
A wheeled autonomous camera robot that follows a person and films them like a professional camera operator. Uses computer vision for person tracking, LiDAR + ultrasonic sensors for obstacle avoidance, and mecanum wheels for omnidirectional movement. Monkey-shaped body designed for emotional attachment.
Computer VisionLiDARROS2Mecanum WheelsReact Native App4K Camera3-Axis Gimbal
AgentOps — Multi-Agent AI DevOps Platform
Kyma AI Assistant
Completed — Live on GitHub
A multi-domain agentic AI platform with a 5-node LangGraph pipeline, supervisor agent with dynamic routing, self-correcting loop, human-in-the-loop approval, RAG via ChromaDB, and LangFuse observability. Domain-agnostic architecture — swappable knowledge base for any field.
FastAPILangGraphLangChainLangFuseOllamaKind (Kubernetes)Docker ComposeSQLiteChroma
Vocaa Robot — Language Learning Humanoid
Concept Stage
Concept for a low-budget desktop humanoid robot connected to the Vocaa language practice app. Users learn languages by talking directly to a physical robot companion.
HumanoidLanguage Learning
Production Apps — Shipped & Live
Vocaa — Real-Time Language Practice
Speak. Connect. Grow.
Live on Google Play — 177 Countries
Real-time language speaking practice with real humans. Press one button, get matched in under 2 seconds, practice for 5 minutes. AI pronunciation feedback powered by GPT-4o and Whisper.
React NativeNestJSPostgreSQLRedisLiveKit WebRTCOpenAIRailwaySupabase
Milap — Anonymous Social Discovery
Connect Anonymously. Discover Genuinely.
Google Play Closed Testing
Anonymous matching app inspired by an ancient Indian ritual. Progressive identity reveal. WebSocket-based real-time matching with cosine similarity scoring.
React NativeNestJSPostgreSQLWebSocketAI Matching
Skills
Technical Competencies
Depth over breadth. Each bar shows where I am: studying, hands-on, applied in a real project, or a completed curriculum.
Deep learning to agentic systems, with a focus on edge deployment.
- Deep LearningCompleted
ANN, CNN, RNN, LSTM, GRU, Transformer architecture — full curriculum
- NLPApplied
Traditional + modern NLP; BERT, GPT-2, Opus-MT via Hugging Face
- Reinforcement LearningCompleted
Q-Learning, DQN, Policy Gradient, Actor-Critic, RLHF
- Generative AIApplied
Production-grade RAG with ChromaDB; LangChain (Models, Prompts, Chains, Memory, Agents)
- Agentic AIApplied
ReAct, multi-agent (supervisor + workers), MCP, function calling
- Edge AIHands-on
Quantization (PTQ, QAT, GPTQ, AWQ, GGUF) and pruning
- Computer VisionStudying
Prof. Shree K. Nayar's First Principles of CV — edge detection, gradients, sensor pipeline, 3D vision
Learning
Self-Directed Research Tracks
Structured, self-driven curricula alongside my degree — each with a written roadmap and notes.
Quantum Computing
100%13-week Qiskit curriculum (up to Grover's Algorithm), saved to GitHub.
- Gates & Bloch sphere
- Bell / GHZ states & teleportation
- Deutsch → Simon's
- Grover's
Deep Learning / RL / Agentic AI
100%Complete curriculum finished — 83 pages of handwritten-style notes.
- DL architectures
- Reinforcement learning + RLHF
- RAG & LangChain
- Multi-agent & MCP
ROS2 Robotics
40%5-phase curriculum mapped to Articulated Robotics plus Lynch & Park, Thrun, and Siciliano.
- Workspace setup
- Custom Python nodes
- Hardware ↔ software stack
Computer Vision
25%6-layer PhD roadmap, from classical vision to embodied AI.
- Classical CV
- Deep Learning CV
- 3D / Geometric
- Multimodal
- Generative
- Embodied AI
Space Technology
15%6-phase, 24-week roadmap from orbital mechanics to an AI-quantum-in-space capstone.
- Orbital mechanics
- Spacecraft systems
- Capstone: AI + quantum in space
Key textbooks being studied
Modern Robotics
Lynch & Park
Probabilistic Robotics
Thrun
Robotics
Siciliano
Computer Vision
Szeliski
Multiple View Geometry
Hartley & Zisserman
Education
Education
MSc Artificial Intelligence → Robotics Pathway
Berlin School of Business and Innovation (BSBI/UCA)
February 2026 intake · pathway changed to Robotics after 1st semester
- Self-studying AI/ML/DL/GenAI/Agentic AI independently alongside the programme.
- Coursework: credit card default prediction ML pipeline; predictive maintenance (LSTM vs TCN); robotics literature review; AI ethics (healthcare, policing, autonomous weapons).
B.Tech Electrical and Electronics Engineering
Cochin University of Science and Technology (CUSAT), Kerala, India
Foundation: circuit design, embedded systems, signal processing, motor control, power electronics
- This EEE background is a direct advantage for hardware-adjacent robotics research.
Speaking
20+ talks on Technology, AI, and Innovation
Universities, colleges, and schools — grounded in what it actually takes to build in the real world.
AI & RoboticsResearch & InnovationCareer & MindsetTechnical Deep-Dives
Available for academic talks, conference presentations, and guest lectures.
Let's advance the science together
Open to PhD positions, research collaborations, visiting researcher opportunities, and industry research roles.
Affective RoboticsEmbodied AIQuantum RoboticsSpace AutonomyHRI
saurabhkumar9493@gmail.comBased in Berlin, Germany. Open to relocating for the right research opportunity.