Skip to content
View damn8daniel's full-sized avatar

Highlights

  • Pro

Block or report damn8daniel

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
damn8daniel/README.md

Daniil Borisov

AI Engineer & Systems Developer

RUDN University — Faculty of Artificial Intelligence | Moscow, Russia

Telegram Email GitHub

77 repositories | 170+ projects | 6 languages | Everything from scratch

I build things from scratch to deeply understand how they work — from GPT-2 and diffusion models to OS kernels and database engines. My repositories span AI/ML, systems programming, and production infrastructure, with 77 public projects containing real, working code.

Tech Stack

Python      ████████████████████████  Expert     │  PyTorch, NumPy, FastAPI, LangChain, CrewAI
C++         ████████████████████░░░░  Advanced   │  C++20, CMake, STL, SIMD, OpenGL
Go          ████████████████░░░░░░░░  Advanced   │  Goroutines, gRPC, net/http, Chi
Rust        ████████████████░░░░░░░░  Advanced   │  Tokio, Axum, async, Serde
TypeScript  ████████████████░░░░░░░░  Advanced   │  Next.js, React, Node.js, Express
Swift       ████████████░░░░░░░░░░░░  Proficient │  SwiftUI, UIKit, MapKit, CoreML

AI Engineering Hub — 93+ Projects

ai-engineering-hub — A comprehensive resource for learning and building with AI. Full roadmap from Python basics to production AI systems.

Category Count Highlights
Beginner 22 OCR (Llama, Gemma, Qwen), Chat UIs, Basic RAG, Image Generation
Intermediate 48 AI Agents (CrewAI, AutoGen), Voice Bots, MCP Servers, Model Benchmarks
Advanced 23 Fine-tuning (Unsloth, GRPO), LangGraph, NotebookLM Clone, Transformers from Scratch

Key technologies: LLMs (Llama 4, DeepSeek-R1, Qwen3) | RAG (LlamaIndex, LangChain, ChromaDB, Milvus) | Agents (CrewAI, AutoGen, LangGraph) | MCP Protocol | Voice (Whisper, AssemblyAI)

Deep Learning — Architectures from Scratch

Project What it does
gpt2-from-scratch GPT-2: multi-head causal attention, KV-cache, BPE tokenization, nucleus sampling
ddpm-from-scratch Diffusion model: UNet backbone, cosine noise schedule, DDIM sampling, EMA
vae-gan VAE-GAN hybrid: PatchGAN + VGG perceptual loss, latent interpolation
nerf-from-scratch NeRF: hierarchical volume rendering, positional encoding, novel view synthesis
clip-from-scratch CLIP: contrastive image-text learning, ViT encoder, zero-shot classification
flow-matching Flow Matching: optimal transport, rectified flows, classifier-free guidance
moe-transformer Mixture of Experts: sparse routing, load balancing, shared experts, RoPE, GQA
vqvae-image-gen VQ-VAE: EMA codebook, PixelCNN autoregressive prior, two-stage generation
mamba-ssm Mamba/S4: selective scan, HiPPO initialization, O(n) sequence modeling
image-captioning ViT encoder + GPT decoder, cross-attention, beam search, BLEU/CIDEr metrics
asr-russian Russian ASR: Conformer encoder, CTC/Attention hybrid, SpecAugment, beam search + LM
time-series-transformer Informer + PatchTST: ProbSparse attention, multi-horizon forecasting

Deep Learning — Training & Optimization

Project What it does
dl-framework-python PyTorch-like framework from scratch: autograd, Conv2d, BatchNorm, Adam, Transformer
distributed-training-framework Distributed training: DDP, FSDP, Pipeline Parallel, Ray Tune hyperparameter search
rlhf-pipeline RLHF: Bradley-Terry reward model, PPO with KL penalty, 3-stage alignment
knowledge-distillation KD: Hinton, FitNets, Attention Transfer, Deep Mutual Learning. ResNet & BERT compression
contrastive-learning SimCLR + BYOL: NT-Xent loss, LARS optimizer, linear eval, KNN probing, t-SNE
lora-from-scratch LoRA/QLoRA: low-rank adaptation, NF4 quantization, adapter composition
sparse-autoencoder Mechanistic interpretability: feature extraction, dead neuron resampling

ML Foundations — From Scratch

Project What it does
autograd-engine Mini PyTorch: computational graph, automatic differentiation in pure NumPy
gradient-boosting-from-scratch XGBoost-like: Newton boosting, histogram splits, L1/L2 reg. Beats sklearn
bayesian-neural-network BNN: Bayes by Backprop, MC Dropout, SWAG, Deep Ensembles, OOD detection
neural-ode Neural ODE: adjoint method, latent ODE, continuous normalizing flows, RK45
svm-from-scratch SVM: SMO algorithm, kernel trick (RBF, poly), soft-margin, SVR, multi-class
em-algorithm EM: GMM, HMM (Baum-Welch), Mixture of Factor Analyzers, BIC/AIC selection
gnn-from-scratch GNN: GCN, GAT, GraphSAGE, GIN. Node/graph classification, link prediction
anomaly-detection Multi-method: Autoencoder, VAE, Isolation Forest — all from scratch

Reinforcement Learning

Project What it does
alpha-zero AlphaZero: MCTS + dual-head ResNet, self-play, ELO tracking. TicTacToe/Connect4/Othello
world-model DreamerV3: RSSM, imagination-based planning, symlog predictions
multi-agent-rl MAPPO: cooperative multi-agent, centralized training decentralized execution
meta-learning MAML, Reptile, ProtoNet: few-shot classification, bilevel optimization
federated-learning FedAvg, FedProx, differential privacy, non-IID partitioning

AI & ML — Production

Project What it does
neural-architecture-search NAS: DARTS, evolutionary, Bayesian optimization with GP, Pareto front viz
on-device-stt On-device STT: Conformer + CTC, mel spectrogram, INT8 quantization, RU + EN
llm-red-teaming Red-teaming: 322 attack prompts, jailbreak/bias/toxicity detection, EU AI Act
data-drift-detection Drift detection: PSI, KL, KS, CUSUM, ADWIN — all from scratch. Auto-retrain
multi-agent-code-review Multi-agent code review: 71+ rules, bug/security/style/arch analyzers
skin-cancer-ensemble 23-model ensemble (SwinV2, ConvNeXt V2, ViT, ResNet-152D). AUC ~0.96
yolo11-waveguide-segmentation YOLO11-seg for space waveguide induction soldering. CLAHE preprocessing
fakehunter-review-detection Fake review detection: fine-tuned RuBERT, F1 ~99.9%, FastAPI backend
skin-cancer-detector Medical AI: EfficientNetB3 + Flask + React, multi-language
monte-carlo-risk-analyzer Monte Carlo simulation: PERT, triangular, lognormal. HTML reports

NLP & Language

Project What it does
nlp-pipeline Production NLP: BPE tokenizer, BiLSTM-CRF NER, Word2Vec, text classification
recommendation-system RecSys: collaborative filtering, matrix factorization, A/B testing framework
lang-interpreter Programming language: lexer, parser, AST, classes, closures, GC

Systems Programming — C++

Project What it does
database-engine-cpp SQL database: B+ tree, WAL, buffer pool, query planner, transactions
redis-clone-cpp Redis clone: RESP protocol, skip list, RDB/AOF persistence, 800K ops/sec
network-stack-cpp TCP/IP stack: Ethernet → ARP → IP → ICMP → UDP → TCP from scratch
ray-tracer-cpp Path tracer: BVH (SAH), Monte Carlo GI, PBR materials, Perlin noise
mini-os-cpp x86-64 kernel: paging, process management, round-robin scheduler, VGA shell
game-engine-cpp 2D engine: ECS architecture, OpenGL renderer, physics, particle system
cpp-json-parser Streaming JSON parser: recursive descent, 1GB+ files, ~50MB memory
cpp-telegram-bot Telegram bot in C++20: games, calendar, reminders, diet tracking

Systems Programming — Go & Rust

Project What it does
api-gateway-go API Gateway: load balancing, circuit breaker, rate limiting, health checks
observability-platform Full observability: logs + metrics + traces, ClickHouse, gRPC collectors
mini-container-runtime Container runtime: namespaces, cgroups v2, OCI image extraction
event-driven-rust Event-driven microservices: NATS, saga pattern, OpenTelemetry
search-engine-rust Search engine: inverted index, BM25 ranking, phrase queries

Backend & Infrastructure

Project What it does
ml-serving-platform ML serving: FastAPI, Celery, Redis, Prometheus, model versioning
realtime-data-pipeline Streaming pipeline: Kafka, TimescaleDB, WebSocket dashboard
ai-saas-platform AI SaaS: Next.js, Stripe billing, OAuth, usage tracking, RAG
k8s-ml-orchestrator MLOps: Helm charts, Argo Workflows, canary deployments, GPU scheduling
workflow-engine DAG workflow engine: topological scheduling, Celery workers, retry logic
async-web-framework Web framework on raw asyncio: routing, middleware, DI, OpenAPI gen
video-streaming-platform Video platform: HLS/DASH, FFmpeg transcoding, adaptive bitrate
identity-server OAuth2/OIDC: PKCE, JWKS rotation, MFA (TOTP), token introspection
graphql-federation-platform GraphQL Federation: Apollo Gateway, 4 subgraph services
crdt-collaboration-engine Real-time collaboration: CRDT, Yjs, conflict resolution, Monaco Editor

Mobile & Apps

Project What it does
brushely-ios iOS dental health: brushing timer, tooth map, camera tracking, SwiftUI
parkspot-ios P2P parking rental: MapKit, booking system, user profiles, SwiftUI + MVVM
pixel-pet-tamagotchi Telegram WebApp virtual pet: FastAPI + aiogram, stat system, SQLite
airdraw Air drawing: MediaPipe hand tracking, gesture recognition, Python + Swift

Popular repositories Loading

  1. brushely-ios brushely-ios Public

    iOS dental health app built with SwiftUI. Brushing timer, interactive tooth map, camera preview, hand motion tracking, particle effects. MVVM architecture.

    Swift 1

  2. damn8daniel damn8daniel Public

    Config files for my GitHub profile.

  3. skin-cancer-ensemble skin-cancer-ensemble Public

    23-model ensemble for skin cancer detection (SwinV2, ConvNeXt V2, EfficientNet V2, ViT, ResNet-152D). AUC ~0.96 on HAM10000. Gradio UI, FastAPI, Docker, Kubernetes.

    Python

  4. yolo11-waveguide-segmentation yolo11-waveguide-segmentation Public

    YOLO11-seg for automatic segmentation of induction soldering stages in space waveguide production. Multi-model comparison, real-time inference, CLAHE preprocessing.

    Jupyter Notebook

  5. fakehunter-review-detection fakehunter-review-detection Public

    Fake review detection using fine-tuned RuBERT (cointegrated/rubert-tiny2). FastAPI backend with single/batch/file inference. 99.9% F1 score.

    Python

  6. airdraw airdraw Public

    Draw in the air using hand gestures captured by camera. MediaPipe hand tracking with gesture recognition. Python + native macOS Swift app.

    Swift