Devesh Tiwari

I build AI systems
and research infra.

Co-Founder & COO @ Buddy2Buddy — building useful AI-native products. I’m drawn to language-model training, speech, and the infrastructure that makes intelligent systems work in the real world.

Building from Mumbai, India.
Co-founded Credarix to turn AI research into production-ready startup software.
Researching domain-adapted speech recognition for specialised technical vocabulary.
Llm infrastructure scaling

Short bio

I’m a product-minded AI engineer and builder based in India. I work across research experiments, agentic workflows, backend systems, and full-stack applications—usually at the point where an AI idea needs to become useful software.

I care about understanding systems all the way down, then explaining and sharing what I learn in public.

Career

I started by shipping software for early-stage products, then moved toward AI-native work: building agentic systems, LLM workflows, inference experiences, and backend infrastructure. Along the way, I co-founded Credarix, contributed to AI products across education and fitness, and grew a campus community around applied ML.

EXPERIENCE

Work experience

Jan 2026 – PresentCo-Founder & COOBuddy2Buddy

Building and operating an early-stage product as a co-founder.

Jun 2026 – Jun 2026Software EngineerMitrah Technologies

I worked on varied projects currently live on playstore 1.Rankbooster which is a educational app for jee and Neet preparation 2. Biker dotcom it is a client delivery app which has two apps one client side and one for bikers 3.mitrah billing it is a all in one solution for end to end software workflows with automations I worked with the AI part in all of them Automation and workflows

AI-Native Apps

Jul 2025 – Jul 2026AIML-Club headSLRTCE

• Led the AIML Club at Shree L. R. Tiwari College of Engineering, fostering a collaborative environment for students interested in AI and machine learning. • Organized technical events and workshops, providing hands-on sessions on reinforcement learning, transformers, and AI architecture.

Jul 2025 – Jul 2025Full stack AI developerVisionX-Technologies

Project Prose: Engineered a multi-agent PPT generation pipeline using LangGraph with state-persistence. Project OGSocial: Developed real-time Node.js backend APIs and AI logic for multimedia social platforms.

Agentic AINextjs

Jun 2025 – Jul 2025AI EngineerConfidence App InstaPrepsAi

Designed and deployed a custom interface for training and inference of "Confidence Models" from AWS . Technical Win: Optimized the inference pipeline, directly increasing model accuracy and reducing latency for educational users. link-->https://instapreps.com/

AI finetuningmlops

Jun 2024 – Jan 2025Full Stack Ai DeveloperArthlete

It is a early stage startup where I was contributing to their AI Native Product for fitness App it is currently live at playstore

Langgraph