AI / GenAI Engineer

Esat
Akkaşoğlu

LLM, RAG and agentic systems in production — three years of applied AI, two of them shipping GenAI that public institutions depend on.

Currently AI & Software contractor to the CircularAI consortium (EU Horizon Europe), serving 18 public organisations in 8 countries. Evaluation-driven, cost-aware, and comfortable owning a system from retrieval architecture to compliance constraints.

RemoteIndependent contractor · EU & US invoicingFull CET overlap

Hire me Download CV GitHub LinkedIn

0.83

RAG answer F1
from a 0.590 BERTScore baseline

4.3%

Hallucination rate
grounding + citation gates

>99%

Source traceability
zero broken links

0.9s

P50 production latency
P95 4.7s under cost control

01 · Technical skills

Core

RAG pipelinesLangChain / LangGraphAgentic orchestrationMulti-intent routingFunction calling / tool useModel Context ProtocolLLM fine-tuningEmbeddings & vector searchElasticsearch kNN · FAISS · pgvectorEvaluation-driven developmentTracing & observabilityToken & cost optimisationPython · FastAPI · async REST

Also

Gemini · OpenAI · Claude · Azure OpenAIDocker · CI/CD · Linux · AzureTypeScript / ReactGDPR & EU AI Act practiceComputer visionpandasJIRA · distributed agile

Multilingual hardening across 7 languages — 0.92 semantic consistency, 0.96 reproducibility, bias parity within σ 3%.

02 · Experience

AI & Software Contractor — CircularAI Consortium

Nov 2023 — Present

CircularPSP, EU Horizon Europe · engaged via Innciso Innovation SL · remote

Re-architected the RAG pipeline to F1 0.83 (precision 0.94 / recall 0.75) with a 4.3% hallucination rate, re-indexing Elasticsearch from text-based to ID-based vector structures across three release cycles.
Shipped four production AI features — procurement-criteria generation, circular quality control, knowledge hub and funding identification — now used by 18 public organisations across 8 countries.
Built the agentic layer: LangChain/LangGraph tools with multi-intent routing, async FastAPI endpoints and scrapers aggregating global procurement and funding data.

Research Scholar, AI & Software — TÜBİTAK BİLGEM YTE

Nov 2023 — Jul 2025

Türkiye's national informatics & information-security R&D centre · Gebze

Drove CircularAI from prototype to a production-approved architecture through structured LLM/RAG experiments.
Benchmarked chunking, embedding and retrieval strategies; the selected design carried directly into the production build.

Software Engineer Intern, Computer Vision — Istanbul University

Mar 2022 — Sep 2023

Informatics Department · Istanbul

Replaced a manual image-analysis workflow with automated Python computer-vision pipelines over two internship terms.

03 · Projects & achievements

Multi-agent · Python + React/Tauri

FinCoach

An AI personal-finance coach that turns raw financial documents into a risk-scored portfolio review — document extraction, portfolio and Turkish-fund analysis, then risk-aware coaching, chained with tool calling and structured output parsing.

RAG · fine-tuned Gemini

AstroBot

A Discord assistant for astronomy Q&A built with Ebubekir: curated knowledge base, expert-labelled evaluation set, and eight months of retrieval-quality iteration.

Competitions

Top 2% of 579 teams

BTK Akademi Datathon 2026 — 12th of 579 teams, feature engineering + gradient-boosting ensembling
Anadolu Hackathon 2026 — 3rd place (Yandex Türkiye × Sivas University) for real-time delivery route optimisation

04 · Education & publications

B.Sc. Astronomy & Space Sciences

Istanbul University · 2019 — 2026 (expected)
Computational track: Python, statistical modelling, large-catalogue analysis (GAIA).

Two peer-reviewed papers

Turkish Journal of Astronomy and Astrophysics, 2023 — eclipsing-binary minima and pulsating-variable maxima from the IST40 Telescope. ORCID 0009-0001-4036-930X.

Certifications & languages

BTK Akademi — Python & TensorFlow · Amateur UAV pilot licence (SHGM, 2024).
Turkish (native) · English (professional working proficiency).

EsatAkkaşoğlu

Technical skills

Experience