Looking for a job

Михаил Яснов

@0008600

ML / Computer Vision Engineer at DSSL who cut fire and smoke detector false positives by ~20% without losing true positives.

Russian Federation

Message

At DSSL, I rebuilt fire and smoke detection for surveillance cameras, reducing false positives by ~20% with no loss of true positives. I also optimized GPU inference to more than double throughput without noticeable quality loss and shipped the system to the product.

At Intelion Cloud, I built and launched a BTC trading system to production and developed invoice OCR and internal RAG/LLM systems. My work also includes computer vision and multimodal ML for chemistry at Infochemistry Scientific Center, ITMO University, where I co-authored two Q1 papers.

Experience

Work history, roles, and key accomplishments

Current

AI Technical Expert

Current

DSSL

Dec 2025 - Present (10 months)

Rebuilt fire and smoke detection for surveillance cameras with TensorRT optimization and multi-frame event confirmation, reducing false positives by ~20%. Deployed self-hosted Qwen model for AI coding agents and wrote internal course on agentic coding.

TensorRT Onnx Polygraphy GroundingDINO SAM CVAT MLFlow Prometheus Grafana Qwen RAG LLM Inference Optimization Video Analytics Auto Labelling Synthetic Data

CV/ML Engineer

Intelion Cloud

Nov 2024 - Dec 2025 (1 year 1 month)

Developed OCR pipeline for invoices mapped to customer forms in 1C accounting system. Built and launched BTC trading system with anomaly detection and ferroalloy price forecasting, plus waste-bin fill-level estimation with route optimization.

OCR RAG LLM VictoriaMetrics Resume Route Optimization Genetic Algorithms Simulated Annealing Heuristics

Computer Vision Engineer

AI University

Mar 2024 - Nov 2024 (8 months)

Implemented customer-facing detection, multi-object and multi-camera tracking, and face recognition using YOLOv8-10, OpenMMLab, ByteTrack, and InsightFace. Optimized inference with TensorRT, ONNX, quantization, pruning, and distillation for high-load video analytics on Kafka.

YOLOv8 YOLOv9 YOLOv10 OpenMMLab ByteTrack InsightFace TensorRT Onnx Quantization Pruning Distillation Kafka Video Analytics

Data Scientist

Infochemistry Scientific Center, ITMO University

Jun 2021 - Mar 2024 (2 years 9 months)

Applied CV and multimodal ML for chemistry, including OCR of 2D molecular images to SMILES, segmentation, detection, tracking, and GNN. Co-authored two Q1 papers.

OCR Segmentation Détection Trackings GNN Multimodal ML Chemistry