FIG_001
FFmpeg, Inside Out
A deep technical guide to understanding FFmpeg's architecture, multimedia foundations, filtergraphs, memory management, and hardware acceleration.
FFmpegVideoAudio
2026-05-1120 min
BACKAryan Sakhala
MY LEARNINGS
Technical deep-dives, architecture patterns, and lessons learned from building production systems.
ENTRIES09 ITEMS
FIG_002
QAT Engine: Hardware Crypto Acceleration with OpenSSL & Qdrant
Building and deploying Intel QAT Engine for hardware-accelerated TLS. From C source patches and OpenSSL provider architecture to Rustls async benchmarks — achieving 58K+ RSA signs/sec with QAT offload.
QATOpenSSLHardware Acceleration
2026-03-1835 min
FIG_003
Computer Vision Workload Analysis: Memory Leaks & System Debugging
Deep investigation into production CV benchmark failures — tracing memory leaks from 5000-frame tensor replication, fixing Celery visibility timeout infinite loops, and eliminating oneDNN log disk exhaustion.
Computer VisionDebuggingDocker
2026-02-2225 min
FIG_004
RunPod: GPU Cloud Deployment Mental Model
A practitioner's guide to RunPod GPU infrastructure — from Template→Pod→Endpoint architecture to debugging CUDA-compat library conflicts and Docker entrypoint overrides for vLLM deployment.
RunPodGPU CloudDocker
2026-01-288 min
FIG_005
SEAL: Homomorphic Encryption from Scratch
A complete guide to homomorphic encryption using Microsoft SEAL. Learn Ring-LWE security, SEAL parameters (poly_modulus, coeff_modulus, plain_modulus), scale factors, and BFV vs CKKS schemes.
Homomorphic EncryptionSEALRing-LWE
2026-01-2525 min
FIG_006
Intel AMX: Understanding CPU-Based Matrix Acceleration
A deep exploration of Intel's Advanced Matrix Extensions. How tile-based operations differ from vector processing, the role of BF16, and how oneDNN bridges hardware and software for AI inference.
Intel AMXCPUMatrix Computation
2026-01-2315 min
FIG_007
Contextual Retrieval with VLLM and LlamaIndex
A comprehensive guide to implementing advanced contextual retrieval using VLLM for efficient LLM inference and LlamaIndex for robust data indexing. Solve the context loss problem in RAG systems.
RAGLLMVLLM
2026-01-2512 min
FIG_008
Celery: Distributed Task Queues at Scale
Mastering asynchronous task processing with Celery, Redis, and RabbitMQ for high-throughput distributed systems.
CeleryPythonDistributed Systems
2026-01-2210 min
FIG_009
ECS + Fargate: Serverless Container Orchestration
Deploying and scaling containerized applications on AWS with ECS Fargate for zero-infrastructure management.
AWSECSFargate
2026-01-249 min