Hello, I am Ivan (Vanya) Poliakov, R&D ML Engineer specializing in high-performance AI integrations.
A core strength of mine is algorithmic optimization, as evident by my profile on Codeforces where I consistently rank in the top 10% of problem solvers globally.
I bring a strong background in statistics, data analysis, and machine learning, alongside hands-on experience building and deploying AI systems for business applications such as social media automation and e-commerce.
My Recent Work:
- Tweet suggestions for a 300k+ follower X account:
Built an AI agent using RAG and a custom RecSys pipeline to plan and write daily content, with a Slack feedback loop. - 24/7 DM Sales Automation for 1M+ follower vendors:
Stabilized a failing infrastructure by migrating workflows to self-hosted n8n and LangChain for an early-stage startup.
I am currently looking for my next freelance project. If you are a recruiter or a founder - I would be happy to chat :)
Pet Projects
With a group of other students, fine-tuned YOLOv6 object detection model to detect the cups and track the ball from the overhead camera. Implemented robotic arm control with adaptive aim correction that sends commands to the arm through an Arduino chip.
The renderer loads an .obj file, a texture file, and paints a scene into a .tga file. It supports basic lighting and camera movement, z-buffer, Gouraud and Phong shading models.
Pentomino tetris setup with, in which you can play yourself or run an autoplay bot. Autoplay was trained with a genetic algorithm.
A 3D image-stitching tool that aligns two overlapping photos (e.g., from a panorama) by estimating their homography transform with robust regression and warping one of the images onto the other.
My Tech Profile
Languages: C++, Python
Performance Engineering: High-Performance Computing, Planning & Scheduling, Search, Big Data
Applied AI & Machine Learning
Retrieval & Agents: RAG Architectures, Vector Search, Agentic Workflows & Memory, Multi-Modal Embeddings
Deep Learning & GenAI: PyTorch Custom Heads, LoRA, RecSys, Vision (YOLO, SAM-2, Qwen, CLIP), Speech (Whisper, SepFormer)
Frameworks: PyTorch, LangChain, n8n, pgvector, OpenCV, scikit-learn, pandas, NumPy, SciPy
System Architecture & Backend
Architecture: Event-Driven Systems, REST/WebSockets, Rate Limiting, Cloud Cost Optimization, LLM Tokenomics
Data & Infrastructure: PostgreSQL, MongoDB, Redis, FastAPI, Docker, GCP, AWS, CI/CD
Web & Prototyping: TypeScript, React.js, Next.js, Streamlit, Gradio
Education
MSc in Artificial Intelligence at Maastricht University, 2024-2026
- Work: In fall 2024 I started offering services as a freelance ML Engineer.
- Competitions: In fall 2024 I formed the first team in the history of my university to go to ICPC NWERC, quarter-finals of arguably the most prestigious programming competition in the world.
- Research: Currently laying out proposal for my MSc thesis.
BSc in Data Science at Maastricht University, 2019-2023
- Internship: In fall of 2021 I developed a similarity search model for a legal referencing company Garant.
- Thesis: My BSc thesis was about pretraining a conditional StyleGAN model from scratch on limited resources.
I offer the following services:
- Custom RAG solutions: Multimodal, optimized chunking, long-term temporal memory
- AI Integrations & Agentic Workflows: Stateful agents, model inference
- System Optimization: Reducing latency, processing time, and costs in complex pipelines
- Model Fine-Tuning: LoRA, custom PyTorch heads for text, image, and speech models
- Data Consulting: Statistical analysis to validate business strategies