Illia K.
Lead AI Architect
Experience
May 2025 - Present
6 monthsGermany
Lead AI Architect
Predictores.ai
- Designed the Agentic System powering the Predictive Brand Intelligence (PBI) and Predictive Communication Intelligence (PCI) platforms, leveraging LLMs, RAG (embedding + chunking) and tool-use orchestration.
- Drove end-to-end architecture and MLops: containerized services, CI/CD, experiment tracking, and production monitoring in cloud environments (Azure).
- Partnered with product & data teams to define guardrails and privacy-first data flows.
- Implemented a K-Nearest Neighbors model, integrated as part of the Agentic System.
Nov 2022 - Apr 2025
2 years 6 monthsUnited States
Senior AI Engineer
Soul AI
- Led the development of a real-time bidding optimization system using reinforcement learning and predictive analytics to dynamically adjust bidding strategies across Google, Meta and TikTok to increase campaign ROI and reduce wasted spend.
- Integrated Agentic AI to create autonomous agents that dynamically adjust ad creative and bidding strategies, reducing manual intervention and improving cross-platform ad campaign efficiency.
- Architected a serverless, high-throughput backend with AWS Lambda, DynamoDB, Kafka and Kubernetes to handle millions of ad interactions per second with low latency.
- Created an automated content generation platform integrating GPT-based NLP models for text and Stable Diffusion for images, streamlining ad production and enhancing content quality.
- Applied transfer learning techniques to adapt and optimize GPT-3.5/4 and diffusion models using Hugging Face Accelerate for scalable, mixed-precision training across multi-GPU systems.
- Operationalized fine-tuned models on AWS SageMaker by configuring Inference Pipelines and Multi-Model Endpoints for scalable, real-time ad generation and dynamic bidding optimization.
- Managed a unified API system to integrate multiple advertising platforms such as Google Ads and Facebook, improving efficiency and reliability of ad campaign deployments.
- Developed a user-friendly analytics dashboard with React and D3.js to visualize real-time performance data, enabling clients to make data-driven decisions on the fly.
- Automated CI/CD pipelines with Docker, Kubernetes, and Jenkins, reducing deployment time and ensuring smooth integration of new features.
- Mentored three junior engineers on best practices for scalable model fine-tuning, production-grade ML deployment to SageMaker, and high-availability cloud architecture.
- Collaborated with campaign managers to identify time-consuming tasks and turned feedback into specifications for autonomous AI agents, optimizing ad creatives and budgets automatically.
- Led collaborative sessions with marketing, product, and engineering teams to ensure the AI-driven platform met evolving business needs.
Mar 2018 - Nov 2022
4 years 9 monthsUnited Kingdom
Senior AI Engineer
AllHere Education
- Shaped the technical direction of a rapidly growing education technology startup by collaborating with founders and product managers to define platform vision and roadmap.
- Developed an OpenAI GPT-3.5 powered AI system integrated with Retrieval-Augmented Generation (RAG) to provide analytical responses to natural language queries using LangChain, OpenAI embeddings, and Pinecone for semantic search across multimodal data sources.
- Developed and deployed NLP-driven educational tools, including a GPT-3.5-powered chatbot for semantic search and real-time information retrieval, improving student engagement and learning outcomes.
- Built an OCR pipeline using Tesseract for text extraction from scanned PDFs and images, enabling efficient digitization and information retrieval from diverse educational materials.
- Engineered a multimodal content extraction pipeline for textual documents (PDF, Docx) and videos using PyPDF2, docx2txt, spaCy, and Whisper model to integrate diverse data sources into a unified knowledge base for RAG processing.
- Architected a cloud-native microservices infrastructure with AWS Lambda to support scalable communication systems for educational institutions.
- Utilized Snowflake for advanced data warehousing to manage complex raw data, optimize query performance, and enable scalable AI-driven data processing.
- Leveraged Databricks and big data technologies for distributed model training and processing large-scale education datasets.
- Contributed to front-end development of a real-time analytics dashboard using Next.js and TypeScript to integrate multimodal AI insights for student progress tracking.
- Built AI-driven solutions to automate lesson plan and educational content creation, reducing manual efforts for teachers.
- Deployed an AI autograding system using OpenCV and TensorFlow for handwritten assessments and image-based submissions to improve grading speed and accuracy.
- Led fine-tuning of LLM models to improve chatbot responses for more accurate and contextually relevant interactions.
- Streamlined CI/CD processes using GitHub Actions and Kubernetes to reduce operational overhead and improve deployment speed.
Jan 2016 - Feb 2018
2 years 2 monthsUnited States
ML/DL Engineer - Healthcare
CareCloud
- Developed a CNN-based medical imaging platform using PyTorch to analyze X-rays and mammograms, boosting diagnostic accuracy and facilitating informed decisions by healthcare providers.
- Created speech enhancement models for telemedicine platforms using Wave-U-Net to improve transcription quality and communication clarity between physicians and patients.
- Built an OCR engine to transform paper records into structured EHR data, automating data entry, improving accessibility of patient information, and reducing manual entry time.
- Engineered ETL pipelines to process large volumes of unstructured healthcare data stored in Snowflake, ensuring high performance and data quality for ML model training and analytics.
- Designed APIs to integrate electronic health record (EHR) systems with telemedicine platforms, ensuring smooth data flow while maintaining compliance with healthcare regulations.
- Built a scalable database architecture to support FHIR standards, improving speed and efficiency of patient data access for healthcare providers.
- Collaborated with product management to gather requirements and build features aligned with healthcare industry needs.
Languages
English
AdvancedEducation
Oct 2011 - Jun 2015
University of Delaware
Bachelor of Computer Science · Computer Science · Newark, United States
Need a freelancer? Find your match in seconds.
Try FRATCH GPT More actions
Similar Freelancers
Discover other experts with similar qualifications and experience
- English
- Deutsch
2025 © FRATCH.IO GmbH. All rights reserved.