a
anviit22

Anvit

@anviit22

AI Systems Engineer for LLM RAG and Automation

Índia
Inglês, Espanhol, Alemão
Algumas informações são exibidas no idioma inglês.
Sobre mim
I help founders and teams turn messy documents, data, and manual workflows into working AI products. I build RAG assistants, PDF chat tools, LLM powered APIs, workflow automations, FastAPI backends, and deployment ready MVPs. My work usually covers the full build: understanding the use case, designing the flow, connecting APIs, structuring outputs, and preparing the system for real use. I care about clean handover, practical UX, and systems that do not break after the demo. Tech: Python, FastAPI, OpenAI API, LangChain, LlamaIndex, vector databases, Docker, REST APIs, cloud deployment.... Saiba mais

Habilidades

a
anviit22
Anvit
offline • 

Conheça meus serviços

Software e Sites de IA
I will build a document intelligence chatbot for your pdfs, sops and reports
A partir deUS$ 50 /projetoMais detalhes

Portfólio

Experiência profissional

Self_Employed

ML & AI Engineer

Self Employed • Freelance

Sep 2024 - Present • 1 yr 10 mos

ML & AI Engineer specializing in GPU kernel engineering (Triton/CUDA), LLM inference optimization, and production ML systems at Plus91 Technology, Pune. Key work: - Custom Triton decode attention kernel: 2.5× end-to-end speedup on Phi-3 Mini, 28ms TTFT P50, 39.4 tok/s throughput - KV cache compression system (TurboQuant): 4.5× reduction at 100k context via Lloyd-Max quantization - Production LLM serving stack (FastAPI/Redis/PyTorch): 81% cache hit rate - PPO-based RL decision systems: +58% performance vs rule-based baselines - Triton GPU kernels: up to 14.65× speedup over PyTorch baseline