Jayesh Choudhari
Available · Berlin, Germany

Jayesh
Choudhari

AI Engineer & Researcher

Building intelligent systems at the intersection of ML, NLP, Computer Vision, and Agentic AI. Pursuing M.Sc. Cognitive Systems (AI) at Universität Potsdam.

Machine Learning LLMs & RAG Agentic AI Computer Vision NLP
2+
Years Experience
15+
Key Projects
2022
In AI Since
50K+
Lines of code written
Scroll to explore
About Me

"I'm an AI Researcher driven by a small goal: to engineer technology that solves problems and make people's lives a little easier. I focus on healthcare, autonomous systems, research, and also enterprise applications. I'm always motivated by the opportunity to develop AI systems that gives a value & support human progress."

Location Berlin, Germany
Education M.Sc. Cognitive Systems (AI) · Universität Potsdam
Current Focus Areas AI Agents · Statistical ML · Linguistics · Safe AI
Research Interests Autonomous systems · RAG · AI in Healthcare · XAI
Languages English · German · Hindi · Marathi
Selected Work

Top Key Projects

01
Azure Agentic RAG CI/CD
Production multi-tool RAG with Azure CI/CD
02
Explainable Autonomous Driving
XAI + Human-in-the-loop retraining
03
Dial AI
Real-time phone conversational AI
04
AI for Specials
Vision AI for blind & deaf accessibility
Project 01 / 04

Azure Enterprise Agentic RAG CI/CD

A production-ready AI assistant that retrieves, reasons, and responds using a multi-tool agentic pipeline deployed on Azure with automated CI/CD via GitHub Actions.

Azure OpenAI Agentic RAG Azure AI Search GitHub Actions Docker CI/CD
View on GitHub
Architecture
CI/CD PIPELINE RUNTIME AGENT Git Push / PR Feature branch commit GitHub Actions Lint · Test · Build Docker Build & Push Azure Container Registry Deploy to Azure Container Apps / App Service Health Check / Rollback Automated smoke tests User Query API endpoint Agent Orchestrator Multi-tool planning & routing RAG Tool Azure AI Search Function Tool APIs / Actions Azure OpenAI (LLM) Reasoning & synthesis Grounded Response Returned to user Azure Monitor / App Insights Logging · Tracing · Cost tracking
Project 02 / 04

Explainable Autonomous Driving

A post-hoc XAI framework for autonomous driving using EfficientNet B0, and human-in-the-loop learning to predict and audit 5 vehicle actions with continuous feedback retraining.

EfficientNet Grad-CAM PyTorch HITL XAI
View on GitHub
● Ongoing
Architecture
Camera Input Video Feed EfficientNet B0 / B2 Grad-CAM Saliency 5 Vehicle Actions Predicted Accelerate Brake Turn Left Turn Right Hold Human Expert Review Feedback Annotation Retraining Loop Continuous HITL Learning
Project 03 / 04

Dial AI

An end-to-end conversational AI system that speaks to customers in real-time via phone calls. Integrates speech-to-text, retrieves relevant knowledge via RAG, generates intelligent responses with LLMs, and converts them back to natural speech — all with sub-second latency.

Speech-to-Text RAG LLM Text-to-Speech ChromaDB Telephony API
Private
Architecture
Incoming Phone Call Telephony API (Twilio / SIP) Speech-to-Text (STT) Whisper / Deepgram RAG — Knowledge Retrieval ChromaDB · Embedding Search Knowledge Base LLM Response Generation GPT-4 / Local LLM Text-to-Speech (TTS) ElevenLabs / Azure TTS
Project 04 / 04

AI for Specials

An accessibility AI system with two modules: a Vision Transformer-based scene understanding pipeline that narrates surroundings for visually impaired users in real-time, and an audio-to-sign-language translation system that converts spoken audio into sign language gestures for deaf and hard-of-hearing individuals.

Vision Transformer Scene Understanding Audio-to-Sign ASR TTS Accessibility AI
View on GitHub
● Ongoing
Architecture
FOR BLIND FOR DEAF Camera / Image Real-time video feed Vision Transformer (ViT) Scene Classification Object Detection YOLO / DETR Caption Generator LLM scene description Voice Output (TTS) Earpiece narration Microphone / Audio Ambient sound capture ASR — Speech Recognition Whisper / wav2vec2 NLP Intent Classifier Context understanding Sign Language Mapper Text → Gesture sequence Sign Animation UI Real-time visual display Blind User Deaf User
Also Built

More Projects

Agentic Auditing & Cleaning Databases

A multi-agent AI system that audits a raw data lake, designs a cleaning plan, reviews it through a committee (agents + human), generates & validates cleaning code, executes it, and produces a final quality report.

GitHub

Universal Dataset Distillation Pipeline

A no-code Streamlit app for LLM knowledge distillation — annotate any dataset with a large "teacher" model, then evaluate or fine-tune smaller "student" models via zero-shot, few-shot, or full weight updates, all from a single UI.

GitHub

Explainable Pneumonia Detection

Medical AI framework achieving AUC > 0.98, comparing CNNs and Vision Transformers. Grad-CAM faithfulness and concentration metrics build clinician trust in AI-assisted radiology.

GitHub

Research Paper → Code Pipeline

Multi-agent Analyst–Architect–Coder workflow using CrewAI and FAISS that transforms academic PDFs into deployable software/research projects.

GitHub

Agentic GitHub Code Reviewer

Privacy-preserving code review using LangGraph, RAG, and local LLMs where no code leaves your machine. Autonomously identifies bugs, security risks, and performance issues in GitHub repositories.

GitHub

More Projects

Career

Experience
Education

Work Experience
June 2025 – Present
AI Engineer
T.A.S Force · Hamburg, Germany
Building enterprise AI applications using LLMs, AI Agents, RAG pipelines, handling cloud infrastructure, CI/CD pipelines, and workflow automation for enterprise use cases.
May 2024 – Sept 2024
AI Researcher Intern
Rovren · California, United States
Created algorithms for an image processing and computer vision platform. Collaborated with Senior Image Processing Scientists on medical and scientific image analysis.
June 2023 – May 2024
AI Consultant Part-Time
Artbider · London, United Kingdom
Built an end-to-end conversational AI system for real-time customer support - integrating STT, ChromaDB, RAG, GPT, and TTS pipelines. Achieved full customer support automation with reduced hallucinations.
Feb 2023 – June 2023
AI Researcher Intern
Dawn Digitech · India
Developed an automated answer-checking ML system using NLTK, SpaCy, and ML algorithms for semantic analysis, grading, and intelligent feedback generation.
Education
2024 – Present
M.Sc. Cognitive Systems (AI)
Universität Potsdam · Germany
2021 – 2023
Master of Computer Applications (Data Science)
Bharati Vidyapeeth University · India
2018 – 2021
Bachelor of Computer Applications
Savitribai Phule Pune University · India
Technical Arsenal

Skills

AI / ML Frameworks
PyTorchTensorFlowScikit-learnHugging FaceOpenAI SDK
NLP & LLMs
Fine-tuningPrompt EngineeringSemantic AnalysisNLTKSpaCy
Agentic AI & LLM Systems
Multi-Agent SystemsAgent Orchestration Knowledge retrievalRAG
Computer Vision
OpenCVSegmentationObject DetectionCNNsVision TransformersGrad-CAM
MLOps & Cloud
AzureAWSDockerRESTful APIsGitCI/CD
Programming & Devlopment
PythonSQLMongoDBNumPyPandasDjangoFastAPI
Credentials

Certifications

Indian Space Research Organisation (ISRO)(GOV)
AI/ML for Geodata Analysis
August 2024
Expertise in AI-driven geospatial data analysis covering ML/DL architectures (CNN, RNN, RCNN, YOLO). Hands-on with Google Earth Engine and Python for spaceborne data processing and environmental spatial analysis.
Let's Connect

Let's Build
Something
Intelligent.

Open to research collaborations, AI/ML roles, and interesting projects.
Based in Berlin - available globally.

Get In Touch