I build Turkish AI.
End to end.
I'm Göktuğ. I turn raw data into trained models, reproducible benchmarks and working inference products. Every claim links back to an artifact you can inspect, run and measure.
DUSUNEN Model Lab.
A measured Turkish AI program: variable-dimension embeddings, compact retrieval, learning-to-rank, pinned benchmarks and working browser inference.
DUSUNEN Atlas 278M
One Turkish retrieval model with six useful embedding sizes. At 128 dimensions it cuts index storage by 83.3% while reaching 94.75% held-out hard-negative accuracy.
DUSUNEN Pusula 118M
A 117.7M encoder with six selectable dimensions, a 91.7% index-size reduction at 32-D, and a pinned five-task Turkish scorecard.
DUSUNEN Mercek 118M
LambdaLoss learning-to-rank over 50K lists, with measured MRR@10 and nDCG@10 gains over its untouched multilingual base on 102,400 frozen pairs.
One protocol. Four systems. Raw evidence.
TurHistQuad, XQuAD, WebFAQ, MKQA and Belebele through the official MTEB evaluator, with pinned revisions and explicit model prompt formats.
50K difficult training triples
Model-mined negatives from a 70K candidate pool, with lexical safeguards, zero fallbacks, checksums and an exact-overlap audit.
Ask in Turkish. Inspect the sources.
A free browser-only retrieval demo that selects extractive evidence and links every answer fragment to its source. No hosted model API and no hidden generation step.
Watch retrieval change from 384-D to 32-D
One browser query, six live Matryoshka vector budgets, instant ranking changes and explicit index-size trade-offs.
NanoSOC1:8B.
A gated Foundation-Sec 8B QLoRA system for structured SOC triage, evidence correlation, MITRE ATT&CK attribution and human-approved response—published with both its benchmark gains and its failure modes.
NanoSOC1 8B · v6 evidence stack
Two task-routed adapters, a 5,492-document security RAG corpus, model-security gates and immutable audit evidence. Built to assist analysts—not to act as an autonomous IDS or response engine.
Inspect the evidence, including failures
A browser-only evaluation companion with frozen holdout metrics, sanitized triage walkthroughs, architecture boundaries and explicit false-positive limitations.
No missing middle.
The work connects data decisions to measured model behavior and finally to something people can use.
Normalize, filter, deduplicate, checksum and publish the exact training split.
Contrastive fine-tuning with memory-aware batches on a single 8 GB GPU.
Retrieve difficult negatives, reject risky matches and record every mining decision.
Pinned held-out data, exact vector search, strong baselines and raw JSON results.
Open weights, documented inference and a live semantic-search experience.