CiteMentor 2.0
Attribution-aware agentic RAG. Pay only for the wisdom you use.
Hybrid RRF
Retrieval Strategy
$0.00
API Cost (local)
Para-level
Attribution
LangGraphHybrid RetrievalRAGMLXChromaDB+2
Read case study →/log
Architect's Log
Production-grade systems, real benchmarks, and the engineering decisions behind them. Every number here is reproducible.
● Complete
Local Invoice Intelligence
81.76% accuracy vs GPT-5.5's 83.04%. 1.28pp gap at zero API cost.
81.76%
Local Accuracy
−1.28pp
vs GPT-5.5
297 docs
Benchmark
OllamaQwen3pdfplumberTesseractPydantic+2
Read case study →● Complete
SLM Hive — Text-to-SQL
53.4% → 77.4%. +24pp absolute. Fully local multi-agent swarm.
+24pp
Accuracy Lift
77.4%
Swarm Accuracy
$2.70
API Cost Saved
LangGraphMulti-AgentOllamaQwen2.5-CoderLlama3.1+2
Read case study →Research
Publications
Peer-reviewed research across NLP, computer vision, and hardware design.
2020Natural Text-to-Speech Synthesis by Conditioning Spectrogram Predictions from Transformer Network on WaveGlow VocoderIEEE – ISCMI2022Table Detection in the Wild: A Novel Diverse Table Detection Dataset and MethodResearchGate2017Gabor Features for Single Sample Face RecognitionIEEE – ICRAECT2016A Comparative Study on Face Recognition using Subspace AnalysisIJCSE2018Design of Energy Efficient and Size Reduced SRAM CellIOSR-JVSP