- π¨βπ» Senior AI/ML Engineer with 6+ years of experience across Machine Learning, Generative AI, Computer Vision, Document AI, and Backend Engineering.
- π€ Building production AI systems involving LLMs, RAG, embeddings, vector search, OCR, document classification, and information extraction.
- βοΈ Experienced in developing scalable services with Python, FastAPI, PostgreSQL, pgvector, MongoDB, Redis, Docker, Kubernetes, and Helm.
- π Hands-on across the complete AI lifecycle: experimentation, evaluation, APIs, CI/CD, deployment, observability, and performance optimization.
- π§ Currently focused on Agentic AI, advanced RAG architectures, AI evaluation, guardrails, and tool-using LLM workflows.
- π MSc in Artificial Intelligence / Informatics from RPTU Kaiserslautern, including a thesis at DFKI on 3D perception and time-to-collision estimation.
- π’ Professional experience across Elevait, Must Analytics, TeamViewer R&D, PwC, and other engineering environments.
- β½ Outside technology, I enjoy football and astronomy.
-
RPTU Kaiserslautern
- Berlin, Germany
- @ajit_kumar_ak
- in/ajit-kumar-ak
Pinned Loading
-
local-rag-system
local-rag-system PublicProduction-oriented RAG and Agentic RAG platform with hybrid retrieval, secure multi-tenant access, custom agent orchestration, MCP business tools, governed actions, feedback loops, CI eval gates, β¦
Python
-
review-intelligence-platform
review-intelligence-platform PublicProduction-style multilingual review intelligence platform for aspect-based sentiment analysis, LLM summarization, temporal insights, fake-review detection, dynamic Q&A, and MLflow-driven evaluation.
Python
-
Thesis_MMdetection3D
Thesis_MMdetection3D PublicThis is my research thesis project titled- Time-to-Collision Estimation for Objects in Autonomous Driving
Python
-
NLP_with_Hugging_Face_Transformers
NLP_with_Hugging_Face_Transformers PublicA project where we see different machine learning algorithms in use with Hugging face pretrained models
Jupyter Notebook
-
DataScience_TimeSeries_HotelLARs
DataScience_TimeSeries_HotelLARs PublicHotel bookings for the Berlin Hospitality Market
Jupyter Notebook
-
Machine_Learning_Concepts_without_libraries
Machine_Learning_Concepts_without_libraries PublicAn important repo where we see basic machine learning concepts without using any pre-built libraries like scikit-learn etc.
Jupyter Notebook
If the problem persists, check the GitHub status page or contact support.