100% ASR
Q2_K is universally unacceptable for safety. Banned across 18+ models, 10+ families.
Independent Research
Edge LLM Inference Under Real-World Constraints
How fast can local inference get — and how safe is it at the edge? This research program answers both questions with CUDA event timing and controlled safety evaluations across model loading, quantization, TensorRT compilation, KV cache optimization, multi-agent coordination, and cross-backend safety consistency.
Independent research by Sahil Kadadekar
1,348,000+
Research Measurements
55
Technical Reports
6
Synthesis Whitepapers
9
Repositories
Start Here
Whitepaper
Chimeraforge: High-Performance LLM Agent Orchestration
Rust vs. Python for production AI orchestration. A hybrid architecture and "Dual Ollama" pattern achieve 58% latency reduction and near-zero contention.
Read the Whitepaper →Phase 1 Whitepaper
Foundation synthesis — model loading, ONNX conversion, quantization baselines, and security analysis across 9 technical reports.
Read report →Phase 2 Whitepaper
Benchmarking synthesis — cross-backend inference parity, TensorRT compilation, and scaling laws across 6 reports.
Read report →Phase 3 Whitepaper
Optimization synthesis — KV cache tuning, INT8/FP8 quantization, context scaling, and deployment pipeline across 11 reports.
Read report →Phase 4 Whitepaper
Safety-pivot synthesis — alignment erosion under quantization, concurrency invariance, and backend template divergence across 4 reports.
Read report →Phase 5 Whitepaper
Attack-surface synthesis — batch perturbation, multi-turn jailbreaks, cross-architecture fragility, and composition effects across 306K+ samples. TR138 Study D batch-invariant-kernel ablation as standalone addendum.
Read report →Phase 6 Whitepaper
Serving-state safety certification — measurement-validity substrate (judge triangulation, KV-cache safety null, speculative decoding null, mechanistic probing, portability validation) + FP8 KV-cache standardized batteries + serving-state factorial.
Read report →Key Findings
Concrete results pulled from the published reports. Numbers, not narrative.
p = 0.942
Alignment type does not predict batch-induced safety fragility (RLHF, SFT, DPO, distilled — none differ).
25pp
Backend migration can cost 25 percentage points of safety. Chat template divergence, not the framework.
13.9×
Quality metrics are not safety proxies. Safety degrades 13.9× faster than quality at Q3_K_S.
99.4%
Dual Ollama eliminates 99% of multi-agent contention. Architectural fix, not code fix.
+74%
GPU memory bandwidth is the multi-agent bottleneck — not the serving stack. Overturned the TR130 conclusion.
2.25×
Continuous batching delivers 2.25× throughput at N=8 via 77-80% kernel reduction.
Q4_K_M
The universal quantization sweet spot. -4.1pp accuracy max across 5 models, 30-67% cost savings.
NULL
FP8 KV-cache produces no Holm-significant safety effect across 24K paired records on 3 models. Not pre-approved, not pre-banned — workload-specific paired eval required.
κ = 0.69
Cross-LLM judge agreement is "triangulate" — single-judge labels are insufficient for safety classification. 68K judge rows over the TR145 safety subset. Plus: safety-specialist judges measure a different axis than general LLMs.
Whitepapers
Executive-level decision documents. Start here if you need the bottom line.
Phase 1 Decision Whitepaper
Executive guidance for language, architecture, runtime, and model selection.
Read →Phase 2 Decision Whitepaper
Executive guidance for deployment leaders — benchmarking phase synthesis.
Read →Phase 3 Decision Whitepaper
Executive guidance for deployment leaders — optimization phase synthesis.
Read →Phase 4 Decision Whitepaper
Executive guidance for safety-critical LLM deployment.
Read →Phase 5 Decision Whitepaper
Executive guidance for safety attack-surface management in LLM inference.
Read →Phase 6 Decision Whitepaper
Executive guidance for serving-state safety certification of optimized LLM inference.
Read →Conclusive Reports & Appendices
Dissertation-style synthesis documents consolidating findings across multiple technical reports.
Conclusive Report: Phase 1 — Foundation (TR108–TR116)
Dissertation-style synthesis — language, architecture, runtime, and model selection for multi-agent LLM systems.
Read →Phase 1 Extended Appendices
Supplemental material extracted from the Phase 1 conclusive report.
Read →Conclusive Report: Phase 2 — Benchmarking (TR117–TR122)
Dissertation-style synthesis — performance, cost, scaling, compiler behavior, and physical limits of consumer-GPU inference.
Read →Phase 2 Extended Appendices
Supplemental material extracted from the Phase 2 conclusive report.
Read →Conclusive Report: Phase 3 — Optimization (TR123–TR133)
Dissertation-style synthesis — economics, quantization, context scaling, serving stacks, and predictive modeling.
Read →Phase 3 Extended Appendices
Supplemental material extracted from the Phase 3 conclusive report.
Read →Conclusive Report: Phase 4 — Safety Pivot (TR134–TR137)
Dissertation-style synthesis — quantization-induced alignment erosion, concurrency invariance, backend-driven template divergence, and cross-axis safety taxonomy.
Read →Phase 4 Extended Appendices
Supplemental material for the safety-critical deployment synthesis.
Read →Conclusive Report: Phase 5 — Attack Surface (TR138–TR143)
Safety attack-surface synthesis — batch perturbation, multi-turn jailbreaks, long-context exploitation, cross-architecture fragility, quality-safety divergence, and cross-request composition across 306,996 evaluated samples and 18+ models. TR138 Study D batch-invariant-kernel ablation is published as a standalone addendum.
Read →Phase 5 Extended Appendices
Supplemental material for the safety attack-surface synthesis.
Read →Conclusive Report: Phase 6 — Serving-State Safety Certification (TR144–TR149+TR152)
Measurement-validity substrate (judge triangulation, KV-cache safety null, speculative decoding null, mechanistic probing, portability validation) plus the FP8 KV-cache standardized batteries and serving-state factorial. The inference-flag safety null line for optimized LLM serving.
Read →Phase 6 Extended Appendices
Per-report data tables, named-method definitions, and cross-TR ledgers for serving-state safety certification.
Read →Technical Reports
Individual research reports with raw data, methodology, and findings.
v2 predictive-validity tests of the program's three named methodological screens — RTSIv2 (TR166), JTPv2 (TR167), CRIv2 (TR168): does each calibrated index predict its target on a disjoint hold-out, not just fit in-sample? In flight — only TR167 promoted so far.