KEDAR SATHE
@wtfkedar
I'm Kedar, a Machine Learning Engineer from Pune. I build end-to-end ML systems, deep learning models, and scalable MLOps infrastructure.
Most of my time goes into model training, optimization, and inference pipelines. As an OSS Maintainer of mermaid-js, and a contributor to projects like vLLM and Weaviate, I enjoy improving serving efficiency and developer tooling.
I also design agentic workflows and retrieval-augmented systems (RAG), and on the systems side, I enjoy low-level programming like custom kernel modules and performance tuning.
Open to ML engineering roles, collaborations, and technical discussions — feel free to reach out.
Work Experience
Machine Learning Engineer
Autonex AI 360
- Fine-tuned and optimized domain-specific LLMs using PEFT techniques (LoRA, QLoRA)
- Integrated OpenCV computer vision algorithms into live camera streams for real-time video processing
- Worked on the internal PM (Project Management) portal to track model training runs and visualization metrics
- Dockerized inference workloads and benchmarked latency across various hardware targets
Jun 2026 – Present
Remote · Mumbai, India
Open-Source Maintainer
Mermaid.js
- Maintains Mermaid.js, a widely used open-source diagramming and visualization library
- Triages issues, reviews pull requests, and guides open-source contributors globally
- Resolved 8+ bugs affecting rendering fidelity and cross-platform compatibility
- Shipped PRs with new functionality and parser optimizations
Jun 2025 – Present
Remote
Software Developer Intern (ML & Backend)
Unique School App India
- Developed a multi-agent RAG system using LangGraph to enable context-aware Q&A over school administrative data
- Designed routing, hybrid retrieval, and response synthesis pipelines for administrators
- Led React Native app migration to integrate AI-powered features and cross-platform UI components
- Set up and improved CI/CD DevOps pipelines for automated backend and API deployment
Jan 2026 – Apr 2026
Pune, India
ML & MLOps Engineer
Freelance
- Built responsive web applications integrated with custom LLMs, vector search, and hybrid RAG systems
- Implemented end-to-end ML training and evaluation pipelines, deploying models on cloud platforms
- Worked with clients to deliver stable, scalable systems containerized with Docker
2023 – Present
Remote
Tech Stack
This list grows faster than my training loss curves — and I love that.
< Machine Learning & Deep Learning />
< MLOps & Infrastructure />
< Languages />
< Data & Databases />
< Frameworks & Web Dev />
< Developer Tools />
Proof of Work
Things I built when curiosity got the better of me.
GroundedRAG
Local multi-agent RAG system built with LangGraph, Ollama, and pgvector. Designed to stay grounded, refuse weak answers, and show its retrieval trail using hybrid search + RRF fusion.
X-Bird
Chrome extension for AI-generated contextual replies on X.com. Supports multiple tones: Auto, Savage, Supportive, Funny, Professional, Casual — with adjustable intensity control.
Blogs & Notes
I write about AI experiments, system design breakdowns, and lessons from building in public. Check out my active blog for technical deep-dives and dev notes.
neuronreads.vercel.app
AI, ML systems, dev notes & more →