# About

I build AI that reads data the way language models read text. My work is Data Language Models, tabular foundation models, and real-time AI systems, taken end to end from research and architecture to production.

.me{float:right;width:230px;margin:0 0 18px 28px}.me img{display:block;width:100%;height:auto;border-radius:12px;border:1px solid rgba(240,242,248,.14)}h2{clear:both}@media(max-width:700px){.me{float:none;width:100%;max-width:300px;margin:0 0 18px}}TransformersAxial attentionPerceiver IOMasked cell modelingContinual learning`Agentic AI`Apache KafkaApache Flink
## Publications
[arXiv:2605.06290Data Language Models: A New Foundation Model Class for Tabular DataThe paper that defines the Data Language Model category: a foundation model that reads raw tables directly.Read →](https://arxiv.org/abs/2605.06290)
## Tabular foundation model: what I built
Schema-2 is the tabular foundation model I built end to end: architecture, pre-training data and objectives, evaluation, serving, and the platform around it. It has more than 1.5 billion parameters and uses axial attention over rows and columns, a Perceiver-style global path, and a schema path. It reads a raw table directly, with no serialization, no preprocessing, and no column names, and one checkpoint of about 2.18 billion parameters carries every capability below. Read the [research](https://www.ozercemkelahmet.com/model/research), the [solutions](https://www.ozercemkelahmet.com/model/solutions), the [architecture](https://www.ozercemkelahmet.com/model/architecture), or the [overview](https://www.ozercemkelahmet.com/model/tabular-foundation-models).
**How to read this list.** Each entry is a benchmark I built a capability for, what the model achieved, and a link to the full story. Scores are left out on purpose: they depend on the harness and the snapshot date, and each story states its protocol. Competitor figures are third-party published values, my figures come from one sealed configuration per benchmark, and I checked every run for contamination.June to July 2026Row-level predictionClassification and regression boards**Achieved.** First on the classification and binary boards at every board size I tested, with one default configuration against rivals that were tuned for hours per dataset. The overall board became valid only after I found and fixed a harness error. Regression and multiclass stayed behind the leaders, and I say so.
[Read how I built it](https://www.ozercemkelahmet.com/model/bench-row-prediction)
XGBoostLightGBMCatBoostTabPFNTabICL13 July 2026Continual learningSequential tasks, no published competitor**Achieved.** The only model in the comparison that takes on new tasks without a full retrain. The recipe that stayed is experience replay with 48 raw rows per task, after I found and fixed an adapter that was being re-created for every task.
[Read how I built it](https://www.ozercemkelahmet.com/model/bench-continual-learning)
Experience replayEWCMIRAS13 July to 3 August 2026Anomaly detectionSemi-supervised ADBench**Achieved.** First among thirteen methods by both average rank and AUROC. Three model-based channels are combined by a label-free pseudo-anomaly selector, so no labels are needed to choose how to fuse the signals.
[Read how I built it](https://www.ozercemkelahmet.com/model/bench-anomaly-detection)
Anomaly detectionIsolation forest17 July to 5 August 2026Probabilistic regressionScoringBench, CRPS and CRLS**Achieved.** The lowest median CRPS and the best mean rank on the merged board of 102 regression datasets and 50 models, from one sealed configuration. A 999-quantile head with split-conformal calibration fixed the coverage weakness, and I re-ran a weakened rival fairly.
[Read how I built it](https://www.ozercemkelahmet.com/model/bench-probabilistic-regression)
18 July 2026Metadata independencePrediction without column names**Achieved.** The only model in the comparison whose accuracy does not fall when column names are removed or replaced with noise. Models that read headers lose roughly seven AUC points when headers vanish.
[Read how I built it](https://www.ozercemkelahmet.com/model/bench-metadata-independence)
18 to 19 July 2026Blind sector identification504 real datasets, no published competitor**Achieved.** A new capability: name the industry sector of an unseen dataset from raw cell values alone, with column names zeroed and no labels. No earlier tabular model performs this task. Pair correlation between columns was the signal that separated neighbouring sectors.
[Read how I built it](https://www.ozercemkelahmet.com/model/bench-sector-identification)
29 July to 2 August 2026Missing-value imputationNumeric and categorical**Achieved.** First place on the real-data board, ahead of language models and the classical imputers in every setting, with the widest margin under not-at-random missingness. Categorical imputation has no published competitor.
[Read how I built it](https://www.ozercemkelahmet.com/model/bench-imputation)
Imputation networksMIDAS31 July to 5 August 2026Entity and schema matchingKeyless and name-blind**Achieved.** Beats the published state of the art on the entity matching benchmarks with no keys, and beats every name-aware matcher on all four schema scenarios while running name-blind.
[Read how I built it](https://www.ozercemkelahmet.com/model/bench-entity-schema-matching)
1 to 2 August 2026Missing-data robustnessPrediction under injected missingness**Achieved.** A mean AUC above the published method across the full grid of 750 settings. It is a narrow lead and I report it as one.
[Read how I built it](https://www.ozercemkelahmet.com/model/bench-missing-data-robustness)
4 to 7 August 2026One checkpoint, then compressionPackaging the whole result**Achieved.** Every capability merged into one unified checkpoint of about 2.18 billion parameters, with the code embedded and one line of bootstrap. The model was dumped at half, 8-bit, and 4-bit precision and checked for fidelity, and the model card was published after a contamination audit.
[Read how I built it](https://www.ozercemkelahmet.com/model/bench-unified-checkpoint)
TurboQuant
## Timeline
The ideas and systems I have worked on, newest first. Orange marks publications, blue marks areas of work. Hover over any tool to see what it is used for, and select it to read more.
2026A new model class for tables: the Data Language ModelTabular foundation model · paper and benchmarksA foundation model that reads raw tables directly, with no serialization and no preprocessing. The paper defines the category. [Read the paper](https://arxiv.org/abs/2605.06290), the [research](https://www.ozercemkelahmet.com/model/research), the [solutions](https://www.ozercemkelahmet.com/model/solutions), and the [benchmark timeline](https://www.ozercemkelahmet.com/model/benchmarks) that shows how each capability was built.
TransformersAxial attentionPerceiver IOMasked cell modelingEWCMIRASTurboQuant2025 to 2026Foundation model for structured dataArchitecture, pre-training, evaluation, serving
- Axial attention over rows and columns, Perceiver IO cross-attention, masked cell modeling for self-supervised pre-training, neural imputation, semantic classification heads, and memory-augmented continual learning.- Large-scale synthetic data, distributed GPU training, mixed precision, evaluation on public benchmarks, and production serving.- Training and inference infrastructure: GPU pipelines, streaming inference, quantization, and compression.- Research on continual learning, catastrophic forgetting, self-supervised objectives for tables, and post-training quantization.
GoPythonNext.jsApache KafkaApache AirflowApache SparkTerraformPrometheusGrafana2024 to 2025Language models, retrieval, and geospatial intelligenceFine-tuning, search, GIS
- Parameter-efficient fine-tuning and domain instruction tuning of language models.- Retrieval over millions of records with hybrid search, vector databases, and cross-encoder reranking.- ML pipelines for classification, prediction, and spatial features, with a GIS and 3D layer for building geometry, setbacks, floor-area ratio, and density.- GPU inference, AI-aware caching, fault-tolerant retrieval, and auto-scaling on AWS.
LoRAQLoRAHybrid searchBM25Cross-encoder rerankingGISAWS2017 to 2024Industrial AI and digital twinsPredictive maintenance, vision, reinforcement learning
- Predictive maintenance with deep learning on IoT telemetry: vibration, thermal profiling, and remaining useful life.- Computer-vision quality control with real-time rejection.- Digital twins backed by physics-informed networks, surrogates, and Bayesian optimization.- Anomaly detection on high-frequency sensor streams with real-time alerting and root-cause analysis.- Reinforcement-learning agents for scheduling, resource allocation, and energy optimization.- LLM assistants and semantic search over maintenance logs, plus big-data pipelines for OEE, forecasting, yield, and energy analytics.
CNNLSTMGRUTransformersAutoencodersVAEOpenCVYOLOMask R-CNNPINNBayesian optimizationIsolation forestOne-class SVMDQNPPOMulti-agent RLApache FlinkOPC UAModbusProfibusMQTTSCADAMESDigital twinsDockerKubernetes2016 to 2017Reactive systems, NLP, and network intelligenceScala, Akka, chatbots, simulation
- Reactive Scala and Akka microservices for fault-tolerant, high-throughput workloads.- NLP chatbots with intent classification and sentiment analysis, plus vector-database retrieval.- Digital twins and neural simulation for predictive maintenance and capacity planning.- Deep learning for image, speech, and video analytics, and reinforcement learning for traffic optimization.
ScalaAkkaSDN/NFVNetwork slicingDigital twinsTransfer learningReinforcement learning2008 to 2016Machine learning on commerce platformsRecommendation, pricing, forecastingMachine learning for recommendations, search relevance, dynamic pricing, demand forecasting, and booking-trend prediction on high-traffic platforms, with real-time integrations, ETL pipelines, CI/CD, caching, load balancing, and BI dashboards.
ASP.NETSQL ServerMicroservicesSOAPSAP HybrisScalaAkka2001 to 2008Foundations: web systems and securityArchitecture, integration, encryptionScalable web systems built on ASP.NET, SOAP, and XML with partner integrations, encryption and secure authentication for user and payment data, and agile project methods.
ASP.NETSOAP
## Focus
Data Language Models and foundation models for structured data, real-time and streaming AI, continual learning, self-supervised objectives for tables, advanced attention, post-training quantization, efficient inference, and agentic systems. My doctoral research is on neural networks, spiking neural networks, and neuromorphic computing.

## AI and machine learning
The methods behind the models I design, train, and evaluate.
Learning paradigmsSupervised learningUnsupervised learningSemi-supervised learningReinforcement learningSelf-supervised learningContrastive learningFederated learningTransfer learningDomain adaptationFew-shot learningZero-shot learningMeta-learningCurriculum learningActive learningReinforcement learningDQNDouble DQNDueling DQNRainbowREINFORCEA2CA3CPPOTRPOSACTD3DDPGMulti-agent RLSelf-supervised methodsSimCLRMoCoBYOLDINOMAEArchitecturesTransformersCNNResNetEfficientNetConvNeXtLSTMGRUGANStyleGANViTSwinGCNGATGraphSAGEDiffusion modelsDDPMDDIMVAEVQ-VAENormalizing flowsNeural ODEMambaS4Mixture-of-ExpertsSpiking neural networksAutoencodersPINNNeuromorphicIntel LoihiIBM TrueNorthSpiNNakerBrainScaleSLavasnnTorchNorseSpikingJellyFoundationsProbabilityStatisticsLinear algebraOptimizationInformation theoryVariational inferenceBayesian deep learningCausal inference
## LLMs, retrieval, and evaluation
Current models, how I adapt them, and how I measure and protect them.
Frontier modelsClaude Fable 5.1Claude Opus 5.5Claude Sonnet 5.5Claude Haiku 5.5GPT-6GPT-5.6Gemini 4Gemini 3.8 FlashGrok 4.7Mistral Large 4Open-weight modelsDeepSeek V4.1Qwen 4Qwen3.8Kimi K3GLM-5.3Gemma 4Fine-tuningFull fine-tuningContinued pre-trainingSFTLoRAQLoRADoRArsLoRAGaLoreAdaptersPrefix tuningPrompt tuningIA3BitFitAlignmentRLHFDPOIPOKTOORPOSimPORLAIFConstitutional AITraining toolingPEFTTRLAxolotlUnslothLLaMA-FactoryTorchtuneDeepSpeed ZeROFSDPMegatron-LMColossal-AIFlashAttentionQuantizationGPTQAWQGGUFbitsandbytesSmoothQuantExLlamaV2TurboQuantRAGNaive RAGAdvanced RAGModular RAGAgentic RAGGraphRAGContextual RetrievalHyDEQuery rewritingHybrid searchBM25Cross-encoder rerankingColBERTRAGASTruLensDeepEvalVector searchPineconeWeaviateChromaDBMilvusQdrantVespaFAISSpgvectorHNSWIVF-PQScaNNDiskANNEvaluation and safetyLangSmithLangfusePhoenixBraintrustPromptFoolm-evaluation-harnessGuardrails AINeMo GuardrailsPresidioRed-teaming
## Agentic AI
Systems where models plan, call tools, and act over many steps.
Agent SDKsClaude Agent SDKOpenAI Agents SDKGoogle ADKMicrosoft Agent FrameworkLangGraphCrewAIPydantic AIMastrasmolagentsLlamaIndexDSPyProtocolsMCPA2AFunction callingStructured outputsOAuth for tool accessAgent patternsReAct loopPlan-and-executeReflectionTree-of-thoughtsOrchestrator-workersMulti-agent systemsHuman-in-the-loopLLM-as-judgeGuardrailsAgent systemsClaude CodeComputer useBrowser agentsLong-running tasksAgent memorySkillsSub-agentsWorkflow automationn8nTemporalCamundaApache AirflowUiPathAgent reliabilityTracingEvalsRegression boardsCost and latency budgetsSandboxingPrompt-injection defense
## Foundation models and applied ML
Building models from scratch, tabular models, and the classical tasks around them.
Foundation models from scratchArchitecture designTokenizer trainingBPESentencePieceSynthetic data generationDeduplicationCurriculum schedulingData parallelismTensor parallelismPipeline parallelismMixed precisionGradient checkpointingTabular foundation modelsAxial attentionPerceiver IOMasked cell modelingImputation networksSemantic sector headsRelated models and baselinesTabPFNTabICLTabSTARFT-TransformerSAINTTabNetXTabContinual learningEWCMIRASSynaptic IntelligenceLwFExperience replaySelf-supervised for tablesDenoising autoencodersInfoNCEMIDASClassical MLscikit-learnXGBoostLightGBMCatBoostRandom forestSVMIsolation forestOne-class SVMPyCaretSHAPLIMEOptunaRay TuneBayesian optimizationForecasting and tasksARIMAProphetN-BEATSTFTPatchTSTChronosAnomaly detectionLearning to rankRecommendationClusteringUMAPVision and multimodalOpenCVYOLOMask R-CNNDETRSAMCLIPBLIP-2LLaVAStable DiffusionControlNetWhisperNeRFGaussian splattingGIS
## Data engineering and lakehouse
How I build governed, streaming, and protected data platforms on AWS and open table formats.
AWS data analyticsAmazon S3AWS Glue Data CatalogAWS Lake FormationAmazon EMREMR on EKSEMR ServerlessAmazon AthenaAWS DataSyncAmazon MSKTable formats and lakehouseApache IcebergIceberg catalogingIceberg compactionSchema evolutionMedallion architectureBronze layerSilver layerGold layerDelta LakeBatch and streaming processingApache SparkPySparkSpark StreamingApache FlinkApache AirflowdbtTrinoEvent streamingApache KafkaKafka ConnectSingle message transformsSchema RegistryMirrorMaker 2MSK ReplicatorEvent-driven architectureRabbitMQNATSData protectionData tokenizationFormat-preserving encryptionProtegrityThalesSpark UDF de-identificationHashiCorp VaultDatabases and syncAmazon RDSPostgreSQLOracle DatabaseChange data captureData synchronizationSnowflakeDatabricksClickHouseMongoDBCassandraRedisElasticsearchDataOpsGitLab CI/CDTerraformAWS CDKSchema-drift detectionData contract validationInfrastructure as codeGovernanceInformatica AxonInformatica EDCData catalogData lineageHow I work with data
- **Cloud analytics.** I work across the AWS data analytics stack: Amazon S3, the Glue Data Catalog, Lake Formation, EMR (on EKS and Serverless), Athena, DataSync, and Amazon MSK.- **Open table formats.** Deep hands-on experience with Apache Iceberg: cataloging, compaction, and schema evolution.- **Processing.** Apache Spark and PySpark, with Spark Streaming for batch, micro-batch, and real-time workloads.- **Medallion lakehouse.** I design and implement Bronze, Silver, and Gold layers end to end.- **Tokenization and encryption at scale.** Protegrity, Thales, format-preserving encryption, and Spark-UDF de-identification pipelines.- **Event-driven architecture.** Kafka and MSK, Kafka Connect with single message transforms, Schema Registry, and MirrorMaker 2 or MSK Replicator.- **Relational databases and sync.** Amazon RDS, PostgreSQL, and Oracle, with data synchronization techniques.- **DataOps.** CI/CD with GitLab, infrastructure as code with Terraform and AWS CDK, schema-drift detection, and data contract validation.- **Governance.** Enterprise data catalogs and governance tools, including Informatica Axon and EDC, and AWS Lake Formation.

## Engineering and infrastructure
Languages, services, serving, cloud, and the operational layer.
LanguagesPythonGoC#C++ScalaTypeScriptRustNode.jsRubySolidityBackendFastAPIFlaskDjangoASP.NET.NET CoreAkkaSQL ServerSAP HybrisSOAPFrontendReactNext.jsReact NativeAngularVue.jsSvelteFlutterArchitectureMicroservicesDomain-driven designCQRSEvent sourcingHexagonal architectureSaga patternCircuit breakerAPIs and securityRESTgRPCGraphQLOpenAPIOAuth 2.0OIDCJWTmTLSZero TrustOWASPGDPRKVKKML frameworksPyTorchPyTorch LightningTensorFlowJAXKerasHugging FaceMLOps and servingMLflowKubeflowWeights & BiasesBentoMLKServeRay ServeTritonvLLMSGLangTensorRTONNX Runtimellama.cppCloud and GPUAWSAzureGCPSageMakerVertex AIKubernetesNVIDIA H100NVIDIA H200NVIDIA B200CUDANCCLCloud TPUDevOps and observabilityDockerHelmIstioArgoCDGitHub ActionsJenkinsPrometheusGrafanaOpenTelemetryELK stackIndustrial IoTOPC UAModbusProfibusMQTTSCADAMESDigital twinsPredictive maintenanceSDN/NFVNetwork slicingTestingpytestJestPlaywrightk6
## Domains
Industries and problem areas I have worked in.
IndustriesSmart manufacturingIndustry 4.0Smart cityUrban intelligenceTelecomFinTechLogisticsE-commerceOnline travelReal-time systemsAugmented reality
## Education and languages

- Integrated doctorate in machine learning and deep learning, Altinbas and Beykent Universities (expected fall 2026).- B.Sc. in Computer Engineering, Eastern Mediterranean University (2003).- Certifications: PMI, PSM (Scrum), and Microsoft MCITP, MCPD, MCP, and MCTS tracks.- Languages: Turkish (native), English (fluent).

## Elsewhere
Find me on [LinkedIn](https://www.linkedin.com/in/ozercemkelahmet), or read the paper on [arXiv](https://arxiv.org/abs/2605.06290). The [articles](https://www.ozercemkelahmet.com/articles) explain ideas in plain language, and the [projects](https://www.ozercemkelahmet.com/projects) are things I have built.

## Contact
Write to [ozer@ozercemkelahmet.com](https://www.ozercemkelahmet.commailto:ozer@ozercemkelahmet.com).

## Key facts
What is a large tabular foundation model?A large transformer pre-trained on a very wide range of tables, so it can predict, fill in missing values, flag anomalies, and match records on a new table in one pass, without training a new model for each dataset.
What is a Data Language Model (DLM)?A Data Language Model is a foundation model that reads raw tables directly, the way a language model reads text, with no serialization, no preprocessing, and no column names. The category is defined in the arXiv paper 2605.06290 by Özer Cem KELAHMET.
What is Schema-2?Schema-2 is the large tabular foundation model built by Özer Cem KELAHMET. It has more than 1.5 billion parameters, uses axial attention over rows and columns, a Perceiver-style global path, and a schema path, and one checkpoint of about 2.18 billion parameters carries every capability.
What can the model do?Row-level prediction, missing-value imputation, anomaly detection, probabilistic regression, continual learning across tasks, entity and schema matching, blind sector identification, and prediction that does not depend on column names.
Who is Özer Cem KELAHMET?A technology and AI leader who builds Data Language Models, large tabular foundation models, and real-time AI systems, end to end from research and architecture to production.
Where can I read the paper?The paper is on arXiv: https://arxiv.org/abs/2605.06290. The documentation on this site covers the architecture, pre-training, evaluation, and every benchmark.