Adithya S Kolavi

AI Researcher, Building Generative AI solutions at Scale

LinkedInX

About

AI Researcher with experience in building generative AI solutions. Skilled in taking products from concept to completion and leading teams effectively. Experienced in AI, ML, and DL technologies, web development, and cloud computing.

Work Experience

CognitiveLab
On Site

January 2023 - present

AI Researcher

Pioneered one of India's first Kannada bilingual Large Language Model - Ambari-7b. Developed Cognitune, an enterprise-grade LLMops platform, reducing time to production by 60%.

TurboML
Remote

June 2023 - present

AI Developer

Contributed to the development of LLMOps platform, focusing on AI system observability and evaluation. Implemented advanced tracing, monitoring, and analytics solutions to gain insights into large-scale AI systems, which enhanced model transparency and performance.

NeoHumans
Remote

June 2023 - present

AI Researcher

Spearheaded the fine-tuning of a Large Language Model (LLM) using a custom synthetic data generation pipeline for multilingual tasks. Successfully deployed the model in a scalable architecture, enabling robust performance across multiple languages and tasks.

Indian Institute of Science (IISC)
On Site

September 2023 - present

Research Intern

Conducting research on Multi-Modal adaptation of LLMs. Training Vision Language model for various use cases such as Video Understanding, Object Detection/Segmentation and Image QnA.

Mandelbulb Technologies
Remote

May 2023 - August 2023

Generative AI Developer

Engineered an end-to-end generative AI application, increasing user engagement by 30% and improving response time by 25%.

Enable Technologies
On Site

September 2022 - December 2022

Full Stack Developer

Developed the e-commerce platform enableupcycling.com, achieving a Lighthouse score of 90 in performance, accessibility, and SEO.

Skills

AI/ML/DL: PyTorch, Transformers, PEFT, Bitsandbytes, Diffusers, Hugging Face Ecosystem
Python Libraries: NLTK, Scapy, FastAPI, Flask, Django, OpenCV, BeautifulSoup, Selenium, Pandas, Poetry, Langchain
Web Frameworks: React.js, Next.js, Express, Node.js, Vue.js, Bootstrap, Tailwind
Cloud: Azure, Azure Machine Learning, AWS, AWS SageMaker, Docker, Kubernetes, Cloudflare, E2E Cloud
Big Data: Databricks, Azure Data Factory, Apache Spark, Hadoop, Kafka
Databases: MongoDB, PostgreSQL, Firebase, Redis, MySQL, Supabase, Pinecone, FAISS, Qdrant, ChromaDb
Languages: HTML, CSS, JavaScript, TypeScript, Python, C/C++, SQL

Research Projects

Indic Eval/Leaderboard

Developed an evaluation framework for Indic Large Language Models, accommodating multiple translated benchmarks and a leaderboard around it for comparison.

spaCy
NLTK
Transformers
SkyPilot
Azure

Ambari-7b

India's first Kannada bilingual LLM utilizing the LLama2/3 base model, fine-tuned across multiple stages with 1 billion Kannada tokens and tokenization efficiency by 85%

Pytroch
Transformers
PEFT
Deepspeed
Azure ML

YoloGemma

Testing and evaluating the capabilities of Vision-Language models (PaliGemma) in performing computer vision tasks such as object detection and segmentation.

LLaVa
PaliGemma
FastGPT
Transformers

VARAG

Vision-Augmented Retrieval and Generation : a system integrating textual and visual information, enhancing RAG by 35% and improving contextual precision by 60%.

LLaVa
Visual RAG
LLama-index
Qdrant

Mixture of Lora Experts

A novel architecture facilitating the dynamic serving of multiple finetuned LLMs by swapping Lora Adapters during inference.

PyTorch
BERT
PEFT
Distributed Training
HPC

ViViD

A state-of-the-art Vision-Language model specialized in converting complex PDFs into markdown with high speed and efficiency.

PyTorch
LLaVA
PEFT
Distributed Training
HPC

Projects

Cognitune

All-in-one platform for LLMops, featuring distributed data processing, multi-GPU fine-tuning, dynamic evaluation, and one-click high-throughput API deployment.

Python
FastAPI
Transformers
Containerization

Storyblocks

Generate Story Video from a Prompt : Transformed text prompts into dynamic story videos with script generation, synchronized audio, and consistent visual style.

Fast API
NextJS
Diffusers
MoviePy
Wisper

Marker API

A production-ready server with 400 github ⭐, easily deployable to convert PDFs, Word documents, etc., into markdown to aid RAG pipelines.

Pytorch
Fast API
HPC
Docker
Transformer

PyRaft

Python implementation of the RAFT consensus algorithm from scratch using FastAPI, achieving a throughput of 50-250 transactions per second

Python
FastAPI
RAFT Consensus

Tokenizer Arena

A friendly arena to easily compare different tokenizers of various LLMs simultaneously, running completely in the browser.

Transformer JS
React
Tailwind CSS

Topic2Dataset

Create high-quality instruction fine-tuning datasets for LLMs by providing a topic or website, allowing massive synthetic data generation.

FastAPI
NextJS
Langchain
GraphScraper
Docker

Education

PES University

2021 - 2025
Bachelor's Degree in Computer Science