ML Engineer (NLP) @ Innopolis University (GPA: 4.88)
Building intelligent systems at the intersection of NLP and production engineering
- Boosted model accuracy by 90% via constrained decoding techniques
- Productionized Hugging Face & VLLM pipelines
- Open-Source Contribution: Added prompt tokens support to VLLM's Logits Processor (#4985)
- Built ETL pipeline: PDF parsing (Nougat) β deduplication β LLM enrichment
- Automated theory generation & structural linking
- Orchestrated 24-GPU cluster (6Γ4xA100) with Prometheus/Grafana monitoring
- Achieved 3x throughput using VLLM with dynamic load balancing
- Reduced manual docs effort by 60% via auto-README generation
- Developed context-aware summarization preserving documentation style
- Architected LLM agent converting NLP queries to API call sequences (action planning)
- Implemented hybrid argument extraction: slot filling + default values
- Integrated with Google Calendar, Todoist, Gmail (+ extensible architecture)
- Developed context preservation system improving planning accuracy by 40%
Languages & Frameworks
Python
PyTorch
Hugging Face
Transformers
VLLM
REST API
ML Engineering
LLM Orchestration
Constrained Decoding
Multi-Agent Systems
NLP Pipelines
Model Optimization
Quantization
Speculative Decoding
Infrastructure
Prometheus
Grafana
RabbitMQ
MLflow
Airflow
Docker
Cluster Management
Load Balancing
GPU Optimization
- LLM agent architecture design
- Production inference optimization (βlatency/βthroughput)
- Multi-modal pipeline development (PDF β structured data)
- Enterprise system integration
- Full-cycle NLP solutions (POC β MVP β Scale)
- Systems Thinker: Design solutions encompassing inference β monitoring (e.g., 24-GPU cluster setup)
- LLM Explorer: Implement cutting-edge techniques from AutoGPT/GPT-Engineer
- Optimization Enthusiast: Experiment with quantization, speculative decoding
BSc in Computer Science
Innopolis University (Expected 2026)
Certifications
- Samsung IT School: Mobile Development (Java, Firebase)
- Cyber-Physical Systems Development (Python, Cloud Tech)
- Endurance kart racing (team competitions)
- Skiing & snowboarding
- Economics & marketing geek
π« Let's connect! LinkedIn | [email protected]