Hi, I'm

Apostolos Dedeloudis

I build production AI systems: LLM routing, model evaluation, inference optimization, and products that ship. I care about measurable outcomes: cost, quality, and systems that work at scale.

Currently building Agentlify, an LLM routing platform that cuts inference cost by 50% with ~5% eval degradation.

Experience

June 2025 · July 2025

Athens, Greece

Market Risk Intern

Alpha Bank

Built and evaluated ML regression models for commodity derivatives pricing, deployed to a production risk pipeline serving €500M+ AUM. Contributed to reporting infrastructure migration.

  • Engineered proprietary pricing models for commodity futures & swaps using Machine Learning and mean-reverting regression models to support internal valuation and satisfy EU requirements for PRIIPs.
  • Quantified exposure on fixed-income portfolios via PFE, xVA and VaR analytics, then sized and back-tested hedges that reduced potential loss by double-digit percentage points.
  • Contributed to the migration of a legacy reporting system onto a new data pipeline, working with front office and compliance to consolidate risk data flows for a €500M+ AUM book.
PythonMachine LearningRegression ModelsVaR / PFE / xVAData PipelinesSQL

June 2024 · July 2024

Athens, Greece

Data & Analytics Intern

KPMG Advisory

Designed and implemented ETL pipelines for enterprise and government data infrastructure. Mapped 25 BI reports across PowerBI, Tableau, and Qlik into clean source-to-target schemas.

  • Planned and implemented SQL ETL pipelines for corporate and public-sector clients.
  • Audited 25 Business Intelligence reports (PowerBI, Tableau, Qlik) and produced source-to-target mappings across 10+ production databases.
  • Delivered client-facing implementation roadmaps and technical blueprints for follow-on projects.
SQLETLPowerBITableauQlikData Modeling

June 2022 · July 2022

Athens, Greece

Software Development Intern

DEMO Pharmaceuticals S.A.

Built internal tooling that digitized business workflows, boosting efficiency by 30%. Designed data pipelines to support ML training for supply chain optimization.

  • Built internal tools (Vue.js/Laravel) that replaced paper-based workflows, improving process efficiency by 30%.
  • Integrated front-end and back-end applications with the company's existing data pipelines.
  • Designed new data pipelines to support ML training for supply chain management.
Vue.jsLaravelJavaScriptData Pipelines

Projects

Agentlify

July 2025

Full-stack platform for configuring and optimizing LLM agents. Uses GMM clustering and a Thompson Sampling multi-armed bandit to route each agentic step to the right model, cutting inference costs by 50% for ~5% eval score degradation.

50%

Cost Reduction

~5%

Drop in Eval Score

PythonFastAPIReactAWSLLMsRLGMMThompson Sampling
Implicit Gender Bias in LLM Hiring

April 2026 · September 2026

My B.Sc. thesis. Measured gender bias in LLM CV screening across 600 matched pairs: implicit cues move scores ~3x more than the name swaps audits test for. Prompting fails to fix it; a decoupled pipeline eliminates it at 1.10-1.36x cost.

~3x

Implicit vs Explicit

600

Matched Pairs

PythonLLM EvalsOpen-weight LLMsStatisticsSLURM / HPC
Aegis

February 2026

Open-source control plane for AI agents. Enforces least-privilege capability policies, builds cryptographic audit logs of every action, and makes every run deterministically replayable for debugging and compliance.

Least-Privilege

Policy Engine

Cryptographic

Audit

GoPythonSQLiteDockergRPC
Flowsage

November 2023 · June 2025

AI diagramming platform I built and ran solo to 15,000 signups and a paying subscriber base. Cuts diagram creation time by ~90%, and still has daily active users a year after I stopped maintaining it.

90%

Speed Improvement

15,000

Signups

AIReactNode.jsLLMs

Education

Bocconi University

B.Sc. in Economics, Management, and Computer Science

September 2023 · November 2026 · GPA: 27.20/30

Pierce, The American College of Greece

International Baccalaureate (IB) & Greek High School Diploma

September 2017 · May 2023 · IB: 43/45

Research

The Portable Text Annotation Tool

University of Michigan · December 2022

Artificial IntelligenceStochastic ModelsQuantitative Finance

Skills

PythonTypeScriptSQLGoDeep LearningReinforcement LearningNLP / LLMsLLM EvalsRAG PipelinesInference OptimizationModel EvaluationData PipelinesAPI DesignAWSDockerReact

Languages

EnglishNative
GreekMother Tongue
FrenchUpper Intermediate
SpanishIntermediate
ItalianLimited Working Proficiency

Courses

Machine Learning · Bocconi University

Big Data & Analytics · Bocconi University

Financial Engineering & Risk Management · Columbia University

Finance & Quantitative Modeling for Analysts · Wharton

AmplifyME Finance Accelerator · Morgan Stanley & UBS

Connect

Let's talk.

I'm always up for a conversation about AI systems, LLM infrastructure, or interesting engineering problems. Drop me a line.

Leadership

Head of Research

Bocconi Statistics and Data Science Association

Co-Founder / Head of Research

Bocconi Students for Quantitative Finance

Co-Founder / Head of IT & Mentors

theHackLab Bocconi Students

© 2026 Apostolos DedeloudisBuilt with Next.js