Full-stack product, applied AI, data engineering
Andreas Andronikakis Full-Stack, AI & Data Engineer
Founding Backend & AI Engineer with 9+ years building software, data, and AI systems across research, high-performance, and fast-growing startup environments. Builds practical, production AI spanning multi-agent platforms, agent orchestration, large-scale data and enrichment pipelines, and cloud-native backends, using Anthropic, OpenAI, Azure, and TypeScript, with hands-on machine-learning evaluation, fine-tuning, and rapid experimentation. Research background in graph databases, large-scale experimentation on noisy, experimental datasets, runtime optimisation, and high-performance computing. Most recently, shipped an AI-assisted document-automation product solo, from architecture and applied AI through to signed release infrastructure and production support.
Current Role
Founder & Sole Engineer at Ktimatologio Pipeline, an independent product built end to end: a Greek land-registry automation platform with document AI, vision-LLM extraction, and a signed Windows distribution.
9+years across research software engineering, backend systems, and managed runtime performance.
6research publications spanning VMs, NUMA systems, Java benchmarks, and data-center architecture.
4EU research programmes supported across INCODE, ENCRYPT, ELEGANT, and ACTiCLOUD workstreams.
AILLMs, prompt engineering, fine-tuning, data curation, deployment, monitoring, and AI-assisted optimisation.
Profile
Systems depth with product pace.
Andreas combines research-grade data engineering with startup execution: designing AI platforms and multi-agent systems, building large-scale data and enrichment pipelines, fine-tuning and evaluating models, shipping a product end to end as sole engineer, and engineering performance from application code down to runtime, memory layout, and hardware architecture.
AI Platforms & Agents
Designs and owns production AI platforms: multi-agent systems, agent orchestration, LLM integration, public APIs, workflow automation, and document AI with verified, self-checking model output, on Anthropic, OpenAI, Azure, and TypeScript.
AI / ML Engineering
Fine-tunes and evaluates models (LoRA, Unsloth, vLLM, Braintrust), curates JSONL datasets, and builds closed-loop, benchmark-driven optimisation systems across Anthropic, OpenAI, and open-source LLMs.
Data & Experimentation
Builds large-scale experimentation and profiling pipelines over graph databases (Neo4j), generating and modelling performance on noisy, experimental datasets across NUMA and heterogeneous hardware.
Experience
Selected work history.
A progression from graph-database and virtual-machine research into large-scale data engineering, AI/ML systems, and founding-stage ownership of a product end to end.
Jul 2026 - Sep 2026
Ktimatologio Pipeline, agentic automation platform (Remote)
Founder & Sole Engineer - Full-Stack, Applied AI & Product
- Shipped an agentic Greek land-registry automation platform as sole engineer; ~29k lines of Node.js and Python whose agents reconcile 1,797 cadastral parcels across 240 client portfolios in a ~4-hour unattended run.
- Built the AI layer: a vision-LLM agent (Anthropic/OpenAI) that takes over when PDF parsing fails, geometrically self-verified against each document's declared area so no unverified output reaches a legal filing.
- Engineered Playwright browser agents that autonomously collect 4,100+ official documents into 2,065 property-right records, reconciled against the source's declared totals to surface 184 silently missing rights.
- Delivered the full product alone: agents propose, humans approve via a multi-tenant 39-route review platform with AES-256-GCM vaults, in a self-contained Windows distribution with Ed25519-signed auto-updates.
Sep 2025 - Jul 2026
PG:AI, Manchester, United Kingdom (Remote)
Founding Software Engineer - Backend & AI
- Sole backend & AI engineer; led PG:AI's rebuild from a CRM product into an AI-native account-intelligence platform.
- Designed a multi-agent architecture on Anthropic, OpenAI, Azure Functions, MongoDB, Xano, and MCP tooling.
- Built AI enrichment pipelines analysing hundreds of thousands of companies, plus a 700+ endpoint public API platform.
- Instrumented model evaluation and observability with Braintrust, Dash0, and PostHog.
Jan 2025 - Jun 2025
University of Manchester, EU Horizon Project INCODE
Research Engineer - AI Systems & Runtime Optimisation
- Designed an AI-assisted optimisation framework for TornadoVM using Java, Python, and large language models.
- Built data-generation pipelines turning benchmark runs into structured JSONL training datasets.
- Evaluated GPT-4o, DeepSeek-R1, Llama, and Mistral, and explored LoRA fine-tuning with Unsloth, vLLM, and Ollama.
- Implemented closed-loop execution of LLM-generated optimisations against measured performance.
Jan 2023 - Dec 2024
University of Manchester, EU Horizon Project INCODE
Research Data Engineer
- Engineered profiling pipelines (Java, Python, Bash) benchmarking Neo4j on a ~100GB LDBC graph dataset across NUMA.
- Ran large-scale benchmarking sweeps generating structured datasets of application- and hardware-level metrics.
- Built a decision-tree classification model predicting NUMA speedup vs. slowdown from profiling features.
- Authored analyses contributing to peer-reviewed publications at ACM SIGPLAN ISMM and MPLR 2023.
Aug 2022 - Dec 2022
University of Manchester, EU Horizon Project ENCRYPT
Research Assistant
- Benchmarked and ran deep data analysis on the Neo4j graph database to identify memory and runtime bottlenecks in managed workloads.
- Built and extended JVM profiling tooling (NUMAProfiler, PerfUtil) on MaxineVM, producing structured datasets for data-driven characterisation.
- Collected and structured large performance datasets supporting architectural tuning and workload profiling.
May 2022 - Jul 2022
University of Manchester, Horizon 2020 ELEGANT
Research Assistant
- Optimised JVM internals and memory/runtime behaviour to improve execution efficiency of the Neo4j graph database on MaxineVM.
- Implemented and evaluated managed-runtime enhancements, generating performance data for graph-database optimisation research.
Mar 2017 - Jun 2017
University of Manchester, Horizon 2020 ACTiCLOUD
Research Software Engineer (Intern)
- Ported the Maxine Virtual Machine to the AArch64 instruction set, enabling graph-database runtime research on ARM-based hardware.
- Debugged and extended the C1X and T1X JIT compilers for the new backend.
- Built automated testing infrastructure in Python and JUnit ensuring correctness and performance across the new platform.
Jan 2016 - Jun 2016
Focus Photography, Chania, Greece
Software Engineer
- Built a Java reporting and performance-analytics platform on a centralised SQL database consolidating company data into productivity metrics.
Capabilities
Technical range.
Practical stack coverage from product engineering to research-grade performance analysis: backend services, data systems, LLM tooling, ML workflows, cloud platforms, and low-level architecture.
Programming
JavaPythonC/C++Go/GolangXanoScriptRC#.NETBashSQLCypherProlog
AI & ML
LLMsGenerative AIPrompt EngineeringOpenAI APIHugging FaceUnslothLoRAPyTorchBraintrustModel MonitoringVision-Language ModelsDocument AIVerified LLM Output
Data Engineering
Neo4jMongoDBMySQLSQLitedbtAirflowMLflowKafkaSparkTerraformPDF Data ExtractionComputational GeometryData-Integrity Reconciliation
Runtime & Architecture
TornadoVMMaxineVMOpenCLCUDANUMAJVM OptimisationCompiler OptimisationGPU/FPGA/CPU
Platforms & Tools
AzureGCPAWSDockerGitHubXano PlatformKongDash0PostHogLinearFigmaNode.jsPlaywrightEd25519 Code SigningRelease Engineering
Product & Research
Technical WritingOpen SourceTeachingEU Research ProjectsBusiness AcumenCross-functional WorkMentoringLaTeXProduct OwnershipEnd-to-End Delivery
Research
Education, publications, teaching.
Academic work focused on memory optimisation, graph databases, managed runtimes, data-center architecture, and hardware/software co-design.
February 2025
M.Phil, Computer Science
University of Manchester. Topic: Performance Analysis for Memory Optimisations of Managed Graph Databases on NUMA Architectures.
2017 - 2025
Ph.D. Candidate, Computer Science
University of Manchester. Research on Neo4j on MaxineVM, NUMA performance analysis, compiler optimisations, GC tuning, and co-design optimisation for graph databases.
2009 - 2017
Diploma, Electronic and Computer Engineering
Technical University of Crete. Diploma thesis: Memory System Evaluation for Disintegrated Cloud Servers. Grade: 7.18/10.
Performance Analysis for Memory Optimisations of Managed Graph Databases on NUMA Architectures
MPhil Thesis, University of Manchester, 2025.
A Multifaceted Memory Analysis of Java Benchmarks
ACM SIGPLAN MPLR, 2023.
Scaling Up Performance of Managed Applications on NUMA Systems
ACM SIGPLAN ISMM, 2023.
dReDBox: A Disaggregated Architectural Perspective for Data Centers
Springer, Hardware Accelerators in Data Centers, 2019.
On the Future of Research VMs: A Hardware/Software Perspective
ACM Programming companion, 2018.
Enabling Virtual Machine Research on ARM-based IoT Devices
Research publication, 2018.
Teaching & Recognition
Academic foundations.
Teaching assistant experience across computer architecture and engineering, plus scholarships and sponsorships from Google, HiPEAC, and the University of Manchester.
Oct 2017 - Jan 2019
COMP15111 Fundamentals of Computer Architecture
Processors, binary representations, fetch-execute cycle, registers, ARM code, data structures, control structures, functions, I/O, and interrupts.
Oct 2018 - Jan 2019
COMP12111 Fundamentals of Computer Engineering
Logic gates, combinatorial and sequential blocks, CAD tools, hardware description languages, and processor implementation.
Feb 2018 - May 2018
COMP25212 System Architecture
Caching, pipelining, multi-threading, multi-core systems, file systems, and virtual machines.
2019
Google Compiler Summit
Full sponsorship to attend the Google Compiler Summit in Munich, Germany.
2018
HiPEAC Summer School
Full scholarship for the Advanced Computer Architecture and Compilation for High-Performance summer school in Fiuggi, Italy.
2017
University of Manchester Scholarship
School scholarship covering Computer Science PhD tuition fees and yearly stipend.