Akshay Parakh

Akshay Parakh

Advisor Data Scientist @ AI, Eli Lilly & Co. | Ph.D. (CSE, IIT Guwahati)

Building single-cell foundation models, in-silico perturbation prediction, and biology-focused reasoning models for drug discovery

Bengaluru, Karnataka, IN
sendtoakshayparekh@gmail.com

About Me

I am an Advisor Data Scientist in the Advanced Intelligence (AI) team at Eli Lilly & Co., Bangalore, where I have worked full-time since June 2022 after joining as an intern in February 2022. My current work focuses on AI for drug discovery: I build single-cell foundation models that capture cellular context, develop in-silico perturbation prediction models across cell types, and explore biological reasoning models, using reinforcement learning and efficient small models. Previously, I led a squad building generative AI for automated first-draft document generation.

I completed my Ph.D. from IIT Guwahati in 2023. My PhD research work centers around Natural Language Processing (NLP), Information Extraction (IE), and Deep Learning (DL), with a primary focus on Large-scale Relation Classification (RC) from noisy datasets.

I enjoy playing Table Tennis & watching Tennis & Footbal. I also like to travel, trek, and explore hidden places and local food. I have keen interest on Finance, Technology and Startups.

Education

Ph.D., Computer Science & Engineering

Indian Institute of Technology, Guwahati.
Advisors: Prof. Ashish Anand & Dr. Amit Awekar (2016-2023)

Diploma in System Software Development(DSSD)

Center for Development of Advanced Computing(CDAC), Hyderabad
(Feb 2014 - Aug 2014)

BE Computer Science & Engineering

G H Raisoni College of Engineering, Nagpur
(2009-2013)

Industry Work Experience

Advisor Data Scientist - Eli Lilly & Co., Bangalore (Oct 2024 - Present)

AI for Drug Discovery: Pre-training foundation models on single-cell data to capture cellular context, and fine-tuning them to predict perturbation effects across cell types, uncovering biological insights from both observed data and model predictions. Additionally, exploring reinforcement learning to build biology-focused reasoning models that understand these computational models and help translate their outputs into biological insight.

Squad Lead for AI Authoring: Using Generative AI, Large Language Models (LLMS), Agentic pipelines, and domain-specific fine-tuning for drafting first draft of manuscripts, reports, presentations, and other documents. Additionally, building intelligent document parser using state-of-the-art research work and paid/open-source tool.

Consultant Data Scientist - Eli Lilly & Co., Bangalore (June 2022 - Sept 2024)

Generative AI: Large Language Models - Fine-Tuning, and Application development (RAG, Question Answering, Summarization, Natural Language Query Parsing); Information Extraction; Semantic Search

Knowledge Graphs: Building Knowledge graph for enterprise data; Leveraging NLG for Information Extraction; Fine-tuning LLMs for Relation Extraction and generating GraphDB queries

NLP Research Intern - Eli Lilly & Co., Bangalore (Feb 2022 - June 2022)

Biomedical Information Retrieval, Unsupervised Key-Phrase Extraction, Transformers, BioClinicalBERT, Biomedical Named Entity Recognition

Research Intern - IBM Research Labs, Bangalore (May 2019 - Aug 2019)

Relation Extraction, Ontology Creation, Natural Language Processing, Deep Learning

Startup Work Experience

Consultant Data Scientist - ClearFeed (Oct 2021 - January 2022)

Natural Language Processing, Deep Learning, Dataset Preprocessing, Text Classification, Summarization & Information Extraction.

Software Developer / Data Engineer - Shopista - Digitales Communication Pvt. Ltd. (Jan 2015 - June 2015)

Android Application UI Designer, Web Scraping & Information Extraction.

Publications and IPs

Patents

  1. Balaji Ganesan, Riddhiman Dasgupta, Akshay Parakh, Hima Patel, Berthold Reinwald. Nueral-Based Ontology Generation and Refinement. [US-11520986-B2]

Conferences

  1. Akshay Parakh, Ashish Anand, Amit Awekar. Taxonomical hierarchy of canonicalized relations from multiple Knowledge Bases. Proceedings of the 7th ACM IKDD CoDS and 25th COMAD.[paper] [resource]
  2. Akshay Parakh, Ashish Anand, Amit Awekar. Improving Relation Classification using Relation Hierarchy.The 27th International Conference on Natural Language & Information Systems, NLDB 2022. [paper]
  3. Abhay M Shalghar, Ayush Kumar, Shobha G, Balaji Ganesan, Aswin Kannan, Akshay Parekh. Document Structure aware Relational Graph Convolutional Networks for Ontology Population.DLG4NLP Workshop at ICLR 2022. [paper]
  4. Akshay Parakh, Ashish Anand, Amit Awekar. Relation Classification Model Performance as a Function of Dataset Label Noise. Proceedings of the 10th ACM IKDD CoDS and 28th COMAD. [paper]

arXiv

  1. Balaji Ganesan, Riddhiman Dasgupta, Akshay Parakh, Hima Patel, Berthold Reinwald. A Neural Architecture for Person Ontology Population.[paper]
  2. Akshay Parakh, Ashish Anand, Amit Awekar. Budget Sensitive Reannotation of Noisy Relation Classification Data Using Label Hierarchy.[paper] resources]
  3. Akshay Parakh, Ashish Anand, Amit Awekar. Noise in Relation Classification Dataset TACRED: Characterization and Reduction. [paper] [resources]

Thesis

  1. Akshay Parakh, Ashish Anand, Amit Awekar. Understanding and Mitigation of Noise in Crowd Sourced Relation Classification Dataset. [link] [github]

Projects, Blogs & Tutorials

Projects

rxie — arXiv Trend Analyser

A Streamlit application that pulls recent papers from selected arXiv category listings, then uses Google Gemini (via LangChain) to produce a structured analysis of research themes, topic clusters, and how they relate to configurable research interests.

Project 2026 Python, Streamlit, LangChain, Gemini Live app · GitHub

Single-Cell Perturbation Prediction

In-silico perturbation prediction on pseudobulk Perturb-seq data, with baseline and ML models for cross-context analysis across K562 and RPE1 datasets.

Project 2026 Python, Perturb-seq, Single-cell ML, Poetry

Blogs

Tutorials

PyTorch Basics: Getting Started with PyTorch

A collection of comprehensive Jupyter notebook tutorials covering the fundamentals of PyTorch. Learn to build neural networks, work with tensors, implement training loops, and develop deep learning models through hands-on examples and practical exercises.

Tutorial 2024 Python, PyTorch, Deep Learning, Jupyter

Generative AI Playground: Getting Started with LangChain & LangGraph

A comprehensive tutorial repository covering the fundamentals of Generative AI development using LangChain and LangGraph. Learn to build RAG applications, multi-agent systems, and advanced AI workflows through hands-on Jupyter notebooks.

Tutorial 2024 Python, LangChain, LangGraph, Jupyter

Model Quantization Tutorial: Getting Started with LLM Quantization

A comprehensive Jupyter notebook tutorial covering the fundamentals of model quantization for Large Language Models (LLMs). Learn to implement quantization techniques, optimize model inference, and reduce memory footprint while maintaining performance through hands-on examples.

Tutorial 2024 Python, PyTorch, Transformers, Model Quantization, Jupyter

Skills

Programming Languages

C & Python

Data Analysis & Visualization

numpy, pandas, matplotlib, & seaborn

ML, DL, & NLP

scikit-learn, pytorch, huggingface-transformers, langchain, gensim, nltk, & spacy

Graph DBMS (Knowledge Graphs)

neo4j & Cypher

Generative AI

Langchain & LangGraph

More on LinkedIn

Languages

  • Hindi: Native Speaker
  • English: Professional Proficiency