Mayank Sharma

About Me

I'm a Senior Applied AI & Data Leader with 5+ years of experience, I've contributed to academic publications (AAAI 2024), built systems serving 10,000+ users daily, and continue to explore the frontiers of GenAI, NLP, and multimodal AI.

M.Tech CSE - Data Science, IIT Jammu · India

My Story

My journey in AI began during my M.Tech in Data Science at IIT Jammu, where I developed a deep fascination for how machines can understand and generate human language. What started as academic curiosity evolved into a career dedicated to making AI systems that are not just powerful, but also trustworthy and practical.

During my time at nference.ai, I transitioned from pure research to building production healthcare systems. I explored everything from CLIP for chest X-rays to large language models for clinical information extraction. This experience taught me that the real challenge isn't just achieving high accuracy in research papers—it's deploying reliable systems that healthcare professionals can trust with patient care.

Today, at Lamipak, I lead AI research initiatives focusing on multimodal RAG and regulatory intelligence. I'm building systems that serve thousands of factory workers daily, combining research rigor with production reliability. Every day, I work on problems that require not just technical expertise, but also deep understanding of user needs and real-world constraints.

Beyond my day job, I'm passionate about knowledge sharing. Through my blog Token by Token, I write about AI, ML, LLMs, RAG systems, and the practical challenges of building GenAI applications. I believe in learning in public and helping others navigate the rapidly evolving AI landscape.

My Journey

From IIT Jammu to production AI systems serving thousands — here's how I bridge the gap between research papers and real-world impact.

Academic Foundation

M.Tech in Data Science from IIT Jammu (CGPA: 8.70/10). Masters thesis on super-resolution in medical imaging. Served as Teaching Assistant, mentoring 100+ students in Algorithms, Software Tools, and Web Engineering.

Medical NLP Research

At nference.ai, transitioned NLP research to production healthcare systems. Explored CLIP for chest X-rays, investigated Meta-OPT and Nvidia Nemo-GPT for clinical extraction, and optimized BERT models for real-time clinical deployment.

Scaling to Production

Now at Lamipak, leading AI research initiatives in multimodal RAG and regulatory intelligence. Building systems that combine research rigor with production reliability, deployed across multiple countries and languages.

Impact & Achievements

AAAI 2024 Publication

Co-authored OpenMedLM (AAAI 2024, NEJM AI), showing that strategic prompt engineering can match or exceed fine-tuning performance in medical question answering using open-source LLMs.

10,000+ Daily Users

Deployed multilingual multimodal RAG system (LamiOps) serving factory workers in production. Measured impact: 99% reduction in manual troubleshooting time, with high retrieval precision and task-completion success rates.

Industry Recognition

Received Bravo Award at nference for optimizing knowledge distillation, achieving 9× inference speedup and 8× model compression while maintaining clinical accuracy in production healthcare systems.

Global Research Impact

Built regulatory intelligence platform analyzing 200+ data sources across 18+ countries and 9+ languages, achieving 95% reduction in manual R&D research workload through automated extraction.

Few-Shot Learning Breakthrough

At nference, demonstrated that few-shot prompting could replace supervised annotation pipelines, reducing reliance on labeled data by 99% while maintaining extraction accuracy using large-scale models (Meta-OPT, Nvidia Nemo-GPT-20B) for clinical information extraction.

Academic Mentorship

Served as Teaching Assistant at IIT Jammu (2019-2021), supporting 100+ students across undergraduate and postgraduate programs in Algorithms, Software Tools, and Web Engineering through tutorials and assignment design.

Beyond the Code

Built on hard work, powered by a touch of talent, sprinkled with humor, and fueled by infinite nerdiness.

Staying Active

When I'm not debugging models, you'll find me balancing code with cardio — gym sessions for strength, running for endurance, and cycling for exploring new trails. Movement keeps the mind sharp and the ideas flowing.

Fledgling Bookworm

A growing passion for reading keeps me curious beyond technical papers. Whether it's exploring new ideas, learning from diverse perspectives, or simply unwinding with a good story, books are becoming an essential part of my routine.

Grounded in Mindfulness

Meditation helps me stay centered amidst the fast-paced world of AI. It's not just about relaxation, it's about clarity, focus, and maintaining perspective when tackling complex problems.

One thing to remember about me: my humility isn't flattery — it's a value shaped by a humble upbringing.

Skills & Expertise

AI/ML Technologies

PyTorch Transformers LangChain ONNX Quantization AWS Bedrock Model Pretraining & Fine Tuning OpenAI API OpenRouter API

Data & Infrastructure

Qdrant FAISS MongoDb SQL PySpark Amazon Web Services (AWS) Docker FastAPI Django Crawl4AI Git

Core Specializations

Natrual Language Processing (NLP) Large Language Models (LLMs) Multimodal RAG Medical NLP Information Extraction Hallucination Reduction Retrieval Evaluation Prompt Engineering