Aman Sahani
Multi-Disciplinary Systems & Engineering Research

Aman Sahani

A System Engineer's journey through AI, machine learning, and innovative software development.

Featured Spotlight

Curated publication from our research index

All Papers
Artificial Intelligence and Machine Learning

SPOTLIGHT PUBLICATION

Fast, Deterministic Decisions at the Edge: Benchmarking TypeSafe Jev vs. GLiNER2.5-Decide on Indian Financial Markets

Aman Sahani• 15 min read

Fast, Deterministic Decisions at the Edge: Benchmarking TypeSafe Jev vs. GLiNER2.5-Decide on Indian Financial Markets

Traditional LLM prompt-and-parse pipelines introduce unacceptable latency (1,500–4,000ms) for high-frequency operational triage. We benchmark non-autoregressive decision models—Fastino's local 340M GLiNER2.5-Decide (50.6ms on CUDA, \$0.00 cost) and TypeSafe Jev (79.0% zero-shot accuracy at \$0.042/1M tokens)—across 500 real-world Indian financial news articles from Hugging Face, exploring GLiNER's advanced multi-head batching, zero-shot entity extraction, and LoRA hot-swapping.

Read Complete Paper

Recent Technical Articles

Latest peer-level writings & architectures

View Archive

8 Core Knowledge Disciplines

Specialized technical taxonomy & research pillars

Engineered Systems & Projects

Production services, compilers, and performance tools

QuantStream: Real-Time Blockchain Mempool Telemetry & EVM Block Builder Engine

Production-grade, distributed real-time blockchain telemetry and block execution engine in Rust with PyTorch training pipeline and Lambda Architecture (sub-millisecond Hot Path AI inference and durable Cold Path persistence).

Custom CUDA & Triton GPU Kernels

High-performance CUDA C++ and Triton JIT kernels for accelerated non-autoregressive decision models, FlashAttention, and zero-allocation memory pooling.

Zero-Allocation Go Microservices

High-concurrency web service engine built on Fiber/fasthttp with memory buffer pooling, lock-free SPMC channels, and OCI cloud deployment.

About Aman Sahani

System Engineer specializing in AI runtime optimization, low-latency microservices, and cross-disciplinary research. I bridge formal mathematical theorems, systems programming, and high-throughput cloud infrastructure to create resilient software engines.

Initiate Collaboration

Interested in discussing distributed architectures, AI kernels, or research inquiries?