Skip to content
View pradeek1120's full-sized avatar
  • Bengaluru, India

Block or report pradeek1120

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.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

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

Report abuse
pradeek1120/README.md
AI Systems, Inference, and Agents

Pradeep K

AI Engineer | Inference Systems | Reliable Agents

I build and evaluate AI systems, from GPU kernel optimization and LLM serving to grounded agent applications. My work focuses on measurable performance, correctness, and production-minded engineering.

LinkedIn · Hugging Face · Email

Bengaluru, India · Open to AI Engineer, Inference Engineer, and LLM Engineer roles

Selected Projects

Inference optimization case studies for MXFP4 GEMM, MoE MXFP4, and Mixed MLA decode. The repo documents quantization-aware dispatch, runtime-path tuning, metadata reuse, correctness checks, and benchmark-driven iteration.

Evidence: Benchmark summary · MXFP4 GEMM · MoE MXFP4 · Mixed MLA

ROCm MI355X Triton FP8 MXFP4 Benchmarking

An assistant for AMD ROCm developers that analyzes profiler output and training metrics, identifies likely GPU bottlenecks, and suggests concrete optimizations.

Evidence: Try the demo · Project README · Benchmarking code · SFT training code

PyTorch ROCm vLLM LoRA DPO Gradio

An OpenEnv benchmark for agents that must handle stale or conflicting knowledge. The agent detects unsupported or outdated answers, repairs its knowledge base, and checks whether the corrected answer is grounded.

Evidence: Try the demo · API docs · Task suite · Inference code

RAG Evaluation OpenEnv Grounding FastAPI

An API-first assistant that turns a natural-language goal into a structured workflow. Agents retrieve context, plan work, create tasks and notes, schedule calendar events, and return results through FastAPI.

Evidence: Project README · Agent workflow · MCP tool definitions

Python FastAPI Google ADK MCP Gemini

Focus Areas

  • Inference and GPU performance: ROCm, AMD accelerators, Triton, vLLM, quantization, kernel profiling
  • LLM applications: retrieval-augmented generation, agent workflows, tool use, model adaptation
  • Reliable delivery: evaluation, grounding, correctness, APIs, containers, and cloud deployment

Technical Skills

Area Skills
Languages Python, C++, C, SQL
ML and systems PyTorch, Transformers, Triton, ROCm, vLLM, LangGraph, Google ADK, MCP
Backend and cloud FastAPI, Docker, Google Cloud, Vertex AI, Cloud Run

Contact

LinkedIn · Hugging Face · Email

Pinned Loading

  1. Multi-Agent-Productivity-Assistant Multi-Agent-Productivity-Assistant Public

    Python

  2. profiloai profiloai Public

    AMD GPU assistant that turns ROCm profiler output into bottleneck diagnoses and actionable code fixes.

    Python

  3. self-healing-rag self-healing-rag Public

    OpenEnv benchmark where AI agents detect hallucinations from stale documents, identify misleading sources, repair the knowledge base, and verify corrected answers through a scored multi-step RAG en…

    Python