Skip to content
View ajaykesarwani's full-sized avatar

Block or report ajaykesarwani

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 is supported. 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
ajaykesarwani/README.md

Hi there, I'm Ajay πŸ‘‹

πŸ’« About Me

πŸ”­ I'm an AI Engineer with 5+ years of production software experience. I build RAG pipelines, multi-agent LLM systems, and β€” most recently β€” compressed an 8B-parameter medical LLM down to 0.6B through structured pruning and knowledge distillation.

🩺 Recent Work: Built and deployed an end-to-end medical vision microservice (ResNet-18), utilizing ONNX Runtime INT8 quantization to achieve ~6x faster inference on CPUs, fully automated via Terraform and CI/CD pipelines.

πŸ› οΈ Open Source: Active contributor to PyTorch (C++ / CUDA / Python), fixing memory safety bugs and improving core tensor operations

πŸ—οΈ Before AI, I spent 5 years building high-availability 4G/5G systems at Ericsson and HCL Technologies (Cisco account) β€” that's where reliability-first engineering became second nature, and it still shapes how I build AI systems today.

πŸŽ“ Education: M.Sc. Computer Science (University of Passau) β€” currently completing a focused Master's in AI Engineering, also at Passau.

🟒 Open to AI Engineer / Full-Stack AI Engineer roles in Europe (onsite or remote).

πŸ’¬ Ask me about RAG evaluation, LangGraph, or model compression.

⚑ Fun fact: I enjoy playing chess.


πŸš€ Featured Projects

Project What it does Stack
dermatology-ai-service End-to-End Medical AI microservice β€” deployed ResNet-18 with ONNX INT8 quantization for 6x faster inference, complete with Dull Razor preprocessing, AWS ECS deployment, CI/CD, and a C# desktop client FastAPI Β· ONNX Β· Docker Β· AWS Β· C#
model_miniaturization Compressed an 8B-parameter medical LLM into a 0.6B triage assistant using structured pruning, knowledge distillation & LoRA β€” evaluated across 4 clinical datasets PyTorch Β· Transformers Β· LoRA
policygpt-rag-docs-qa Production-style RAG Q&A over your own documents β€” hybrid BM25 + dense retrieval, LLM reranking, streaming answers, and an eval harness for recall & faithfulness FastAPI Β· ChromaDB Β· Next.js
ai-career-copilot Multi-agent career assistant β€” parses your resume/GitHub/LinkedIn, ranks live job postings, and drafts tailored resumes & cover letters LangGraph Β· Google ADK Β· Gemini
production-research-copilot Multi-agent research assistant β€” planner/web/RAG/writer agents that turn a query into a cited markdown report FastAPI Β· LangGraph Β· Groq

🌐 Find Me

LinkedIn Portfolio


πŸ’» Tech Stack

AI / LLM Engineering Python PyTorch TensorFlow Keras ONNX OpenCV LangChain LangGraph Google ADK Gemini Groq ChromaDB Scikit-Learn Pandas NumPy

Backend & APIs FastAPI Flask Django Docker PostgreSQL MySQL

Cloud & DevOps AWS Azure Google Cloud Kubernetes Terraform Git

Earlier Stack β€” Enterprise & Client Development Java Spring Boot Hibernate C++ C# .NET


πŸ“Š GitHub Stats

Ajay's GitHub stats

Pinned Loading

  1. model_miniaturization model_miniaturization Public

    Jupyter Notebook

  2. policygpt-rag-docs-qa policygpt-rag-docs-qa Public

    PolicyGPT RAG doc QA

    Python 1

  3. ai-career-copilot ai-career-copilot Public

    Python 2

  4. production-research-copilot production-research-copilot Public

    Python 1

  5. Thesis Thesis Public

    Thesis

    Jupyter Notebook 1

  6. dermatology-ai-service dermatology-ai-service Public

    Python