For employers

Hire this AI Trainer

Sign in or create an account to invite AI Trainers to your job.

LoginCreate account
Invite to Job
K
Kristian Adam

Kristian Adam

Agency
Germany flagSchillingsfürst, Germany
vombat.kiwi
$91.00/hrIntermediate1+SOC 2GDPR

Key Skills

Software

Other
CVATCVAT

Top Subject Matter

No subject matter listed

Top Data Types

3D Sensor
TextText
Computer Code ProgrammingComputer Code Programming

Top Task Types

Function CallingFunction Calling
Data CollectionData Collection
ClassificationClassification
SegmentationSegmentation
Fine-tuningFine-tuning

Company Overview

Vombat UG is a specialized European infrastructure provider and AI architect, dedicated to building sovereign, autonomous intelligence systems for enterprise and ecological applications. Our core business is not manual labeling, but the end-to-end construction of AI networks, edge infrastructure, and managed model services. Our Mission: To empower organizations with fully autonomous, data-sovereign AI ecosystems that bridge the digital and physical worlds. We design, deploy, and operate the complete stack: from heterogeneous sensor networks and unified data layers to fine-tuned, on-demand models. Core Services: AI Infrastructure & Networking: We architect and implement robust hardware/software foundations, including Linux-based IoT sensor networks, Edge AI clusters (Apple Silicon), and Unified MCP Gateways. We consolidate disparate SaaS and legacy systems into a Single Source of Truth (SOT) using technologies like Hasura (GraphQL) to enable real-time data flow. Managed Model Training & Operations: We operate the full training lifecycle. Using our curated data pipelines, we fine-tune sovereign models (e.g., Mistral) either on-demand or via autonomous agent swarms in secure cloud environments (Infomaniak). We deliver the result as a managed API service, providing clients with instant access to specialized intelligence without managing the complexity themselves. Autonomous Agent Orchestration: We deploy intelligent agents that utilize Function Calling to execute real-world actions across all connected platforms—from automating administrative workflows (ERP, Accounting) to controlling physical actuators (HVAC, irrigation, robotics). Data Curation & Fine-Tuning: As part of our infrastructure service, we handle the complex annotation of multi-modal data (3D sensors, audio, text, biological metrics) using scientific ground truths to train models for predictive maintenance and physical understanding. Industries & Expertise: We specialize in Industrial Automation (e.g., fully autonomous skilled-trade operations), AgriTech/Permaculture (complex ecosystem monitoring), and Environmental IoT. Our projects range from zero-touch office automation to biological early-warning systems. Security & Sovereignty: Operating exclusively for European clients, we guarantee 100% data sovereignty. All infrastructure is hosted on ISO-certified, green-energy Swiss servers (Infomaniak). We utilize local-first Edge AI strategies to minimize latency and cloud dependency, ensuring GDPR compliance and immunity from the US CLOUD Act. Why vomba UG? We do not just label data; we build the nervous system of your organization. From the physical sensor to the cloud API, we provide the complete architecture for autonomous operation, delivering actionable intelligence as a secure, managed service.

IntermediateEnglishGerman

Security

Security Overview

Infrastructure & Physical Security: Vombat UG leverages ISO 27001/27017/27018 certified data centers by Infomaniak (Geneva, Switzerland) for all data processing, storage, and model training. Physical security includes biometric access controls, 24/7 CCTV surveillance, and redundant power systems. This ensures 100% data residency in Switzerland, guaranteeing GDPR compliance and exemption from the US CLOUD Act. Cybersecurity Architecture: We operate a zero-trust network architecture with end-to-end encryption (TLS 1.3+ in transit, AES-256 at rest). Client projects are isolated in dedicated VPC to prevent cross-tenant data access. Strict RBAC and MFA are enforced for all system access. All company endpoints utilize enterprise-grade encryption and automated threat detection. Data Handling & Personnel: All employees and contractors are bound by strict Non-Disclosure Agreements and Data Processing Agreements. We adhere to the principle of least privilege, granting access only to necessary data. Regular training on data privacy and social engineering is mandatory. AI-Specific Safeguards: Customer data used for fine-tuning (e.g., Mistral models) is processed in isolated environments and is never used to train public foundation models. We guarantee full data sovereignty and intellectual property protection. Regular internal audits and adherence to our provider's third-party ISO certifications ensure continuous compliance. An incident response plan is in place to address potential breaches within 24 hours.

Security Credentials

SOC 2GDPR

Labeling Experience

Multi-Modal Physical AI for Autonomous Aquaponics & Permaculture Ecosystems

Other3D SensorSegmentationClassification
Project Description: Challenge: Managing a complex regenerative ecosystem (1,000m² permaculture + commercial greenhouse) involving aquaponics, biogas, livestock (fish, poultry, insects), and energy systems requires real-time understanding of intricate biological and physical relations. Manual monitoring is insufficient for preventive care and autonomous optimization. Solution: We are architecting a sovereign Physical AI system running on quantized (4-bit) Apple Silicon M3 Pro hardware at the edge. The system ingests massive streams of heterogeneous sensor data from custom Linux-based actors to learn causal relationships and enable autonomous decision-making. Multi-Modal Sensor Fusion: Data sources include water chemistry (pH, O2, nutrients), climate (temp, humidity, light, energy), soil metrics, and visual feeds. We correlate these with external data (weather, seasons) to train models on preventive anomaly detection (e.g., predicting fish disease before outbreaks, optimizing biogas yield). Autonomous Control & Robotics: The fine-tuned model doesn't just monitor; it actively controls the ecosystem (aeration, feeding, climate vents) and is being prepared for robotic actuators for automated harvesting and maintenance. Comprehensive Biodiversity Tracking: Beyond fish and plants, the system tracks livestock (chickens, ducks) and beneficial insects (bees, ladybugs, ants), analyzing their behavior as bio-indicators for system health. Virtualized Data Interface: Developed an autonomous Data Application Platform that visualizes complex data logs into intuitive insights for non-experts (students, staff), making high-level biology accessible via natural language. Data Strategy: Annotation relies on curated scientific datasets combined with manual expert labeling to ensure biological accuracy. Impact (Ongoing): Creating a self-regulating agricultural organism that minimizes resource waste, prevents biological losses through early intervention, and serves as a live laboratory for sustainable AI-driven farming. The project demonstrates the viability of local, energy-efficient Edge AI for complex environmental control.

Project Description: Challenge: Managing a complex regenerative ecosystem (1,000m² permaculture + commercial greenhouse) involving aquaponics, biogas, livestock (fish, poultry, insects), and energy systems requires real-time understanding of intricate biological and physical relations. Manual monitoring is insufficient for preventive care and autonomous optimization. Solution: We are architecting a sovereign Physical AI system running on quantized (4-bit) Apple Silicon M3 Pro hardware at the edge. The system ingests massive streams of heterogeneous sensor data from custom Linux-based actors to learn causal relationships and enable autonomous decision-making. Multi-Modal Sensor Fusion: Data sources include water chemistry (pH, O2, nutrients), climate (temp, humidity, light, energy), soil metrics, and visual feeds. We correlate these with external data (weather, seasons) to train models on preventive anomaly detection (e.g., predicting fish disease before outbreaks, optimizing biogas yield). Autonomous Control & Robotics: The fine-tuned model doesn't just monitor; it actively controls the ecosystem (aeration, feeding, climate vents) and is being prepared for robotic actuators for automated harvesting and maintenance. Comprehensive Biodiversity Tracking: Beyond fish and plants, the system tracks livestock (chickens, ducks) and beneficial insects (bees, ladybugs, ants), analyzing their behavior as bio-indicators for system health. Virtualized Data Interface: Developed an autonomous Data Application Platform that visualizes complex data logs into intuitive insights for non-experts (students, staff), making high-level biology accessible via natural language. Data Strategy: Annotation relies on curated scientific datasets combined with manual expert labeling to ensure biological accuracy. Impact (Ongoing): Creating a self-regulating agricultural organism that minimizes resource waste, prevents biological losses through early intervention, and serves as a live laboratory for sustainable AI-driven farming. The project demonstrates the viability of local, energy-efficient Edge AI for complex environmental control.

2026 - Present

End-to-End Enterprise Autonomy: Zero-Touch Operations for HVAC Company

OtherComputer Code ProgrammingRLHFClassification
A mid-sized HVAC (SHK) company sought to eliminate manual administrative overhead and automate complex operational workflows from inventory to invoicing. Solution: We architected and deployed a fully autonomous AI ecosystem replacing the entire administrative staff. Core Architecture: Orchestrated all SaaS applications via Hasura to create a unified Real-Time GraphQL data layer. Integrated Mistral Studio models, fine-tuned on proprietary company data (contracts, technical manuals, historical logs). Autonomous Agents: Deployed specialized agents for Inbound Communication, Accounting, Inventory Management, and Fleet Coordination. These agents utilize advanced Function Calling to execute real-world actions: reordering stock based on IoT warehouse data, scheduling technician routes, and generating invoices without human intervention. Edge AI & Hardware: Implemented a sovereign, local-first AI infrastructure using Apple Silicon chips for on-demand model inference, ensuring low latency and data privacy. Custom IoT devices (Linux-based) monitor warehouse stock levels in real-time. User Interface: Enabled natural language control for the remaining supervisor via a custom "Le Chat" interface, allowing high-level oversight without manual data entry. Impact: Achieved 100% automation of back-office processes. The company now operates with zero administrative staff (reduced from full team to 1 supervisor). Inventory is self-managing, vehicles are autonomously dispatched, and financial processes run end-to-end without human touch. This case study proves the viability of full enterprise autonomy using sovereign European AI models.

A mid-sized HVAC (SHK) company sought to eliminate manual administrative overhead and automate complex operational workflows from inventory to invoicing. Solution: We architected and deployed a fully autonomous AI ecosystem replacing the entire administrative staff. Core Architecture: Orchestrated all SaaS applications via Hasura to create a unified Real-Time GraphQL data layer. Integrated Mistral Studio models, fine-tuned on proprietary company data (contracts, technical manuals, historical logs). Autonomous Agents: Deployed specialized agents for Inbound Communication, Accounting, Inventory Management, and Fleet Coordination. These agents utilize advanced Function Calling to execute real-world actions: reordering stock based on IoT warehouse data, scheduling technician routes, and generating invoices without human intervention. Edge AI & Hardware: Implemented a sovereign, local-first AI infrastructure using Apple Silicon chips for on-demand model inference, ensuring low latency and data privacy. Custom IoT devices (Linux-based) monitor warehouse stock levels in real-time. User Interface: Enabled natural language control for the remaining supervisor via a custom "Le Chat" interface, allowing high-level oversight without manual data entry. Impact: Achieved 100% automation of back-office processes. The company now operates with zero administrative staff (reduced from full team to 1 supervisor). Inventory is self-managing, vehicles are autonomously dispatched, and financial processes run end-to-end without human touch. This case study proves the viability of full enterprise autonomy using sovereign European AI models.

2026 - 2026