https://nazmi.tech/https://nazmi.tech/about/https://nazmi.tech/approach/https://nazmi.tech/audit/https://nazmi.tech/blog/https://nazmi.tech/blog/01_offline_teacher_trace_sft/https://nazmi.tech/blog/02_offline_soft_label_kd/https://nazmi.tech/blog/03_student_state_hard_token_correction_sft/https://nazmi.tech/blog/04_enterprise_support_chatbot/https://nazmi.tech/blog/04_on_policy_probability_distillation/https://nazmi.tech/blog/05_multi_agent_rag_dspy_gepa/https://nazmi.tech/blog/05_on_policy_self_distillation/https://nazmi.tech/blog/06_agent_optimizes_agent_memory/https://nazmi.tech/blog/07_smolagents_deep_dive/https://nazmi.tech/blog/08_knowledge_distillation_foundations/https://nazmi.tech/blog/09_context_management_compaction_llms/https://nazmi.tech/blog/10_agent_teams_claude_sdk/https://nazmi.tech/blog/11_text_splitting_techniques/https://nazmi.tech/blog/12_hybridrag_knowledge_graphs/https://nazmi.tech/blog/13_rl_in_llms_thoughts/https://nazmi.tech/blog/14_verifying_llm_outputs/https://nazmi.tech/blog/15_self_rag/https://nazmi.tech/blog/16_medusa_llm_inference/https://nazmi.tech/blog/17_speculative_decoding/https://nazmi.tech/blog/18_reft_fine_tuning/https://nazmi.tech/blog/19_agents_not_enough/https://nazmi.tech/blog/21_knowledge_graphs_gnn/https://nazmi.tech/blog/22_inside_transformers_part_1/https://nazmi.tech/blog/23_inside_transformers_part_2/https://nazmi.tech/blog/24_inside_transformers_part_3/https://nazmi.tech/blog/25_inside_transformers_part_4/https://nazmi.tech/blog/26_inside_transformers_part_5/https://nazmi.tech/blog/27_chain_of_draft_prompting/https://nazmi.tech/blog/28_deepseek_r1_rl_reasoning/https://nazmi.tech/blog/29_ai_scientist/https://nazmi.tech/blog/30_rl_fine_tuning_openai/https://nazmi.tech/blog/31_mixture_memory_experts_lamini/https://nazmi.tech/blog/32_kan_kolmogorov_networks/https://nazmi.tech/blog/33_expert_gate_lifelong_learning/https://nazmi.tech/blog/34_mlops_part_1_experiment_tracking_mlflow/https://nazmi.tech/blog/35_mlops_part_2a_workflow_orchestration_prefect/https://nazmi.tech/blog/36_mlops_part_2b_workflow_orchestration_zenml/https://nazmi.tech/blog/37_mlops_part_3_model_deployment/https://nazmi.tech/blog/38_mlops_part_4a_model_monitoring/https://nazmi.tech/blog/39_mlflow_on_gcp/https://nazmi.tech/blog/40_data_engineering_week_1/https://nazmi.tech/blog/41_data_engineering_week_2a/https://nazmi.tech/blog/42_data_engineering_week_2b/https://nazmi.tech/blog/43_data_engineering_week_3a/https://nazmi.tech/blog/44_data_engineering_week_3b/https://nazmi.tech/blog/45_learning_to_compact_llms/https://nazmi.tech/blog/46_post_hoc_context_compilation_lerim/https://nazmi.tech/blog/47_theory_safety_context_compaction/https://nazmi.tech/blog/48_production_gates_for_ai_agents/https://nazmi.tech/contact/https://nazmi.tech/services/https://nazmi.tech/work/