Islam Kussainov@Whitness00
1
Events attended
1
Submissions made
Kazakhstan
3+ years of experience
About me
I am a machine learning researcher and developer based in Kazakhstan, specializing in the intersection of AI, computer vision, and advanced geospatial analysis. Currently studying, my work focuses on deploying production-ready machine learning pipelines to solve complex environmental and historical challenges. My research background includes executing specialized machine learning projects aimed at ecological monitoring, such as predicting the presence of saksaul trees in the Aral Sea region using custom data masking and robust statistical pipelines. Additionally, I have developed automated geospatial end-to-end workflows for cultural heritage conservation, successfully applying high-resolution UAV imagery, Sentinel-2 multispectral data, and geomorphometric auditing to accurately detect historical burial mounds (kurgans).
š¤ Top Collaborators
š¤ Latest Submissions
GemmaJudge: Gemma Attacks and Judges Itself
Every model update can introduce a new failure. GemmaJudge turns that risk into a repeatable release check: Gemma generates targeted adversarial prompts, the target model responds, and Gemma scores each response from 1 to 5 with reasoning and an evidence span. The shipped CLI accepts ASR and failed-case thresholds, emits a machine-readable JSON summary, and exits non-zero when a CI regression gate fails. Our committed AMD proof runs the official google/gemma-3-4b-it model as attacker and judge and google/gemma-3-1b-it as the target through local OpenAI-compatible vLLM endpoints on an AMD Instinct MI300X VF using ROCm 6.4. In the recorded three-case demonstration, the Gemma judge assigned failure scores to 3/3 cases, producing 100% judge-assigned ASR over that small run. The full multi-stage pipeline took 55.31 seconds, while each individual model request was capped at 25 seconds. Repeated temperature-0 scores were stable, but this measures deterministic score stability, not judge correctness. The public Streamlit application is an evidence dashboard over committed real artifacts rather than a public live-inference endpoint. It lets judges inspect the AMD run, individual attacks, target responses, judge reasoning, evidence spans, fine-tuning results, and a recorded five-target demonstration. GemmaJudge does not claim that adversarial testing is new. Its differentiator is an auditable, open-weight attacker-plus-judge loop with a self-hosted AMD path, case-level evidence, and a release decision. A single model family can share blind spots, so high-stakes use still requires human calibration or an independent second judge. Our ROCm LoRA study is deliberately scoped. On one 56-example validation split, also used for checkpoint selection, the tuned Gemma-3-4B judge improved valid JSON from 89.3% to 100.0% and pass/fail accuracy from 66.1% to 75.0%. This is a small internal validation study, not an independent benchmark or a production-readiness claim.
11 Jul 2026
