Sandro is currently unavailable
Sandro Barnabishvili
Verified Expert in Engineering
Machine Learning Engineer and Developer
Tbilisi, Georgia
Toptal member since September 24, 2018
Bio
Sandro holds a BSc and MSc in Computer Science and brings over 15 years of experience in software engineering and 10+ years in machine learning. His core expertise spans NLP, GenAI, MLOps, Big Data, and algorithm design. Sandro has contributed to both large-scale enterprises like Microsoft and Logitech, as well as fast-paced startups, delivering AI-driven solutions across diverse domains.
Portfolio
OPUS-Technologies GmbH
Python, AI Agents, Large Language Models (LLMs), JavaScript, TypeScript...
PlantingSpace
Python, Julia, PyTorch, Amazon Web Services (AWS), Large Language Models (LLMs)...
Lexitas
Prompt Engineering, Large Language Models (LLMs)...
Experience
- Python - 10 years
- PyTorch - 10 years
- Machine Learning - 10 years
- Natural Language Processing (NLP) - 9 years
- Generative Artificial Intelligence (GenAI) - 7 years
- Apache Spark - 6 years
- Retrieval-augmented Generation (RAG) - 4 years
- Large Language Models (LLMs) - 4 years
Preferred Environment
Jupyter Notebook, Linux, Python, Amazon Web Services (AWS)
The most amazing...
...accomplishment of mine has been leading the development of a critical component for a client's workflow, which positively affected millions of users.
Work Experience
Senior Back-end Engineer (Agentic AI)
2025 - 2026
OPUS-Technologies GmbH
- Built a procurement process automation pipeline using an Agentic AI approach, orchestrating document classification, structured LLM extraction, and multi-stage reconciliation across purchase orders, invoices, and ERP records.
- Implemented an evaluation and monitoring system for LLM-driven extraction, tracking field-level accuracy, consistency, and reliability.
- Provided technical consulting on scalable back-end and Agentic system architecture, influencing design patterns, data modeling, orchestration strategy, and quality assurance frameworks.
Technologies: Python, AI Agents, Large Language Models (LLMs), JavaScript, TypeScript, Asyncio, Pydantic, FastAPI, Streamlit, SQLAlchemy, PostgreSQL, Supabase, React, Agentic AI, Temporal Cloud, Prompt Engineering, Natural Language Processing (NLP), Alembic, SQLModel, Gemini API, CI/CD Pipelines, Machine Learning Operations (MLOps), Sentry, Workflow Automation
Senior ML Engineer/Researcher
2024 - 2025
PlantingSpace
- Built low-level vector search functionality for the existing system, written in Julia.
- Conducted experiments comparing sparse/dense/hybrid semantic search methods using open-source embedding models.
- Contributed to LLM-backed projects, including PDF knowledge extraction, text-to-code parsing, synthetic data generation, and open-source LLM deployment and inference.
Technologies: Python, Julia, PyTorch, Amazon Web Services (AWS), Large Language Models (LLMs), Machine Learning Operations (MLOps), Natural Language Processing (NLP), Retrieval-augmented Generation (RAG), Machine Learning, Prompt Engineering, Hugging Face, Docker, NumPy, Pandas, Weights & Biases, GitLab CI/CD, Generative Artificial Intelligence (GenAI), Linux, Deep Learning, Data Science, Agile Product Delivery, Agile Product Management, LangChain, Software Architecture, Vector Databases, LangChain, Architecture, Ollama, Workflow Automation, Abstract Syntax Trees (AST)
Senior ML Engineer
2023 - 2024
Lexitas
- Created a scalable RUG framework for semantic search and long-document summarization using AWS-based microservice architecture, specifically SageMaker, Lambda, SQS, SNS, OpenSearch, CloudFormation, Docker, etc.
- Worked on multistage prompt engineering for multiple document understanding tasks using recent LLMs, open source (LLaMA, Mistral, MPT, and Falcon), and OpenAI's GPT.
- Contributed to model automatic deployment (AWS CloudFormation, Docker, Bash), CI/CD, testing, and evaluation.
Technologies: Prompt Engineering, Large Language Models (LLMs), Retrieval-augmented Generation (RAG), Amazon SageMaker, Machine Learning Operations (MLOps), PyTorch, Natural Language Processing (NLP), Python, AWS Lambda, AWS CloudFormation, Amazon OpenSearch, Microservices Architecture, Elasticsearch, Docker, Testing, Machine Learning, NumPy, Pandas, LlamaIndex, Hugging Face, Generative Artificial Intelligence (GenAI), Linux, Deep Learning, Data Science, LangChain, Amazon Web Services (AWS), Software Architecture, Vector Databases, LangChain, Architecture, Workflow Automation
Applied Scientist
2021 - 2023
Microsoft
- Tasked to focus on certain areas like prompt-based NLP, question answering/generation, summarization, and information extraction.
- Performed in-depth qualitative and quantitative evaluation of Bing's latest GPT-based ML production pipelines, including building relevant codebase.
- Planned and executed research agendas, conducted ML experiments, participated in brainstorming, and tracked recent papers.
Technologies: Natural Language Processing (NLP), Python, PyTorch, Machine Learning, SQL, Machine Learning Operations (MLOps), NumPy, Pandas, Big Data Architecture, Large Language Models (LLMs), Azure, JavaScript, HTML, Hugging Face, Research, Generative Artificial Intelligence (GenAI), Prompt Engineering, Linux, Deep Learning, Data Science, Software Architecture, Vector Databases, LangChain, Architecture, Workflow Automation
AI Invited Lecturer
2020 - 2022
Free University of Tbilisi
- Taught class "NLP with Deep Learning" for CS students.
- Supervised multiple Bachelor thesis projects in NLP and Recommendation Systems.
- Created course syllabus and prepared practical materials. Planned exam sessions with TAs.
Technologies: Natural Language Processing (NLP), Python, PyTorch, Machine Learning, NumPy, Pandas, Big Data Architecture, Linux, Deep Learning, Data Science, Hugging Face, Generative Artificial Intelligence (GenAI), Vector Databases, LangChain, Architecture, XGBoost, Workflow Automation
Lead Machine Learning Engineer
2018 - 2022
Kyros
- Optimized large‑scale ETL and ML Spark workflows, resulting in huge cost savings and hours into minutes speedup.
- Implemented CI/CD, large‑scale automated tests, model and data versioning (MLFlow), and automatic deployment (Airflow, Docker, Azure).
- Led the development of an AutoML platform for speeding up DL model prototyping, fine‑tuning, and deployment. Features include friendly API, multi‑node distributed training, Bayesian hyper‑parameter search, advanced logging, and analytics.
- Planned new ML-based product features with the CEO, creating new user stories and doing code reviews.
Technologies: Python, Apache Spark, Machine Learning Operations (MLOps), PyTorch, Machine Learning, SQL, NumPy, Pandas, Big Data Architecture, Azure Databricks, Natural Language Processing (NLP), Linux, Deep Learning, Data Science, Agile Product Delivery, Agile Product Management, Docker, Software Architecture, Architecture, XGBoost, Workflow Automation, Databricks, Apache Airflow
AI Consultant
2018 - 2020
MaxinAI
- Acted as the lead researcher on the project about estimating video scene motion and texture complexity for optimal transcoding and gave SOTA results compared to the reference paper.
- Helped the team develop multiple nutritional information extraction algorithms from food product images. Used a combination of OCR, language modeling, object detection, clustering, and graph algorithms.
- Implemented a few computer vision POCs for company clients, including real-time object detection and tracking, image quality enhancement, semantic search, etc.
- Added a PostgreSQL C++ extension to support high-dimensional vector searches.
Technologies: Python, PyTorch, Machine Learning, SQL, Natural Language Processing (NLP), Machine Learning Operations (MLOps), Computer Vision, C++, NumPy, Pandas, Big Data Architecture, Generative Artificial Intelligence (GenAI), Linux, OpenCV, Image Processing, Torchvision, Deep Learning, Data Science, Agile Product Delivery, Agile Product Management, Amazon Web Services (AWS), Hugging Face, Flask, Docker, FastAPI, Software Architecture, Vector Databases, Architecture, XGBoost, Workflow Automation
Machine Learning Engineer
2018 - 2018
Logitech
- Oversaw R&D in end-to-end dialog modeling, long-term memory in neural networks, text representation, and intent understanding.
- Designed and built a software framework for creating dialog assistants, utilizing the latest DL techniques.
- Created a synthetic dialog data generation tool for mitigating data deficiency problems with end-to-end dialog learning.
Technologies: PyTorch, Python, Natural Language Processing (NLP), Machine Learning, SQL, NumPy, Pandas, Big Data Architecture, Deep Learning, Data Science, Generative Artificial Intelligence (GenAI), Docker, Architecture
Software Engineer
2015 - 2016
AlphaCredit
- Designed and implemented raw text document storage and full-text search engine with REST API.
- Created a client-server web application for searching text fragments efficiently.
- Integrated the new solution to the client's legacy software by accessing API from MS SQL database.
- Managed the full development cycle, helped the client to refine their needs, and delegated tasks to the junior developer.
Technologies: SQL, Big Data Architecture, Machine Learning, Natural Language Processing (NLP), Data Science, PHP, Software Architecture, Vector Databases, Architecture
Full-stack Software Engineer
2014 - 2015
Alta Software
- Implemented front-end and core business logic components of an enterprise billing software for SOCAR, one of the biggest oil companies in the region.
- Optimized database access via the ORM framework and SQL queries to improve software performance.
- Implemented BI tool with a financial data dashboard, user management, and data reporting functionalities.
Technologies: SQL, Architecture
Full-stack Web Developer
2010 - 2011
Happy Group
- Created multiple websites with a content management system from scratch using pure PHP.
- Built a content management system framework like WordPress, allowing building small and medium-sized websites very fast.
- Implemented a few JavaScript plug-ins for existing websites.
Technologies: SQL, PHP
Experience
Large-scale Machine Learning for Citation Prediction
A big data R&D project at a university lab where I built a tool for automatically extracting and linking citations to cited resources in a large collection of online news articles, around 100GB scale distributed dataset.
Technologies: Python, Apache Spark, HDFS, NLP, Graph algorithms, linear models.
Technologies: Python, Apache Spark, HDFS, NLP, Graph algorithms, linear models.
Language IDE
I've designed a simple programming language with up to 10 commands, procedures, and recursion, and wrote its interpreter and IDE (Windows desktop application). The work has been used at several educational institutions to teach programming basics and resulted in a scientific publication.
Technologies: C# with WPF
Technologies: C# with WPF
GPU-accelerated Image Patching App
As a university project, I implemented a desktop app (C++, Qt) that used Poisson's partial differential equation-solving parallel algorithm written in pure CUDA and C++ from scratch to blend one image patch into another.
AutoML Framework
I built the ML framework for accelerating the DL model development lifecycle. The product became the main component of the client's entire AI workflow.
Some of the features include:
- Rich high-level API, enabling fast prototyping and deployment.
- Multi-node multi-GPU distributed training in Azure Databricks infrastructure.
- Automatic hyper-parameter search
- Multi-task learning
- Advanced logging and monitoring
- Automatic deployment in the client's production pipeline
Tech stack: Azure Databricks, Apache Spark, Apache Petastorm, PyTorch, MLFlow.
Some of the features include:
- Rich high-level API, enabling fast prototyping and deployment.
- Multi-node multi-GPU distributed training in Azure Databricks infrastructure.
- Automatic hyper-parameter search
- Multi-task learning
- Advanced logging and monitoring
- Automatic deployment in the client's production pipeline
Tech stack: Azure Databricks, Apache Spark, Apache Petastorm, PyTorch, MLFlow.
Agentic Workflow Automation for Procurement Process
AI and back-end engineering for extracting structural information from complex PDF documents and automating a 3-way procurement process using LLM Agentic workflows. Monitoring, logging CI/CI, and experimentation.
Education
2016 - 2018
Master's Degree in Computer Science
Swiss Federal Institute of Technology in Lausanne (EPFL) - Lausanne, Switzerland
2012 - 2016
Bachelor's Degree (with Honors) in Computer Scienece
St. Andrew the First-Called Georgian University - Tbilisi, Georgia
Certifications
APRIL 2019 - APRIL 2022
AWS Certified Solutions Architect - Associate
Amazon Web Services
Skills
Libraries/APIs
PyTorch, NumPy, Pandas, XGBoost, OpenCV, PyTorch Lightning, Asyncio, Python Asyncio, Pydantic, SQLAlchemy, React
Tools
Torchvision, Apache Airflow, Amazon SageMaker, AWS CloudFormation, Amazon OpenSearch, GitLab CI/CD, Sentry
Languages
SQL, Python, C++, Julia, JavaScript, HTML, C#, PHP, TypeScript, Scala
Platforms
Amazon Web Services (AWS), Ollama, Databricks, Docker, Linux, AWS Lambda, Weights & Biases, Azure, Temporal Cloud
Frameworks
Apache Spark, Flask, LlamaIndex, Horovod, Streamlit, Alembic
Paradigms
Microservices Architecture, Testing, Object-oriented Programming (OOP), Unit Testing
Storage
Elasticsearch, PostgreSQL
Other
Machine Learning, Natural Language Processing (NLP), Prompt Engineering, Large Language Models (LLMs), Retrieval-augmented Generation (RAG), Hugging Face, Generative Artificial Intelligence (GenAI), Software Architecture, Deep Learning, Data Science, Vector Databases, Architecture, Workflow Automation, Machine Learning Operations (MLOps), Computer Vision, Big Data Architecture, Azure Databricks, Image Processing, Agile Product Delivery, Agile Product Management, LangChain, FastAPI, LangChain, Abstract Syntax Trees (AST), Research, Algorithms, Agentic AI, AI Agents, Supabase, SQLModel, Gemini API, CI/CD Pipelines
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring