Aahan Singh
Verified Expert in Engineering
Machine Learning Engineer and Developer
Bengaluru, Karnataka, India
Toptal member since October 24, 2022
Aahan is an applied scientist and ML engineer with 7+ years across agentic AI, RAG systems, and foundation models. At G42, he architected LangGraph pipelines, optimized multi-source RAG, and co-authored 3 NeurIPS 2025 papers (Gene42, Prot42, Chem42). Previously built clinical AI at Qritive—100% sensitivity cancer detection, 85.7% latency reduction, full MLOps stack from scratch. He works across the full depth: LLMs, fine-tuning, K8s deployments, and benchmarking. He ships fast and scales clean.
Portfolio
Experience
- Artificial Intelligence (AI) - 6 years
- Python 3 - 6 years
- Agentic AI - 2 years
- AI Agents - 2 years
- LangGraph - 2 years
- Retrieval-augmented Generation (RAG) - 2 years
- Agentic Frameworks - 2 years
- LangChain - 1 year
Preferred Environment
Python 3, Docker, Kubernetes, Machine Learning, Large Language Models (LLMs), LangGraph, Agno, AI Agents, API Integration, Generative Artificial Intelligence (GenAI)
The most amazing...
...project I worked on was a prostate cancer AI module with 100% sensitivity, 95% specificity, cutting radiologist workload by 57%. Deployed clinically.
Work Experience
Senior Machine Learning Engineer | AI Scientist
G42
- Transformed a core RAG (retrieval-augmented generation) architecture by optimizing static pipelines, achieving measurable improvements in retrieval accuracy and system latency.
- Architected and delivered an agentic pipeline using LangGraph and LangChain within a two-week sprint, demonstrating rapid prototyping and technical adaptability.
- Developed a modular prompt engineering framework supporting multi-section system prompts to manage and deploy 17+ distinct AI personas across the organization.
- Integrated dual-retrieval strategies (web and internal index search) with dynamic source whitelisting to ensure robust, grounded responses for multi-source reasoning.
- Enhanced factual grounding and contextual reasoning by refining agent logic, resulting in highly objective and balanced AI-generated outputs and integrated graph and table generation tools, enabling real-time data visualization within the chat interface.
- Built a unified, production-ready back end supporting both chat and voice-to-text modalities, successfully deployed for high-level executive decision-support tools.
- Automated end-to-end deployments of AI agent services to Kubernetes clusters using Helm, significantly reducing deployment overhead and increasing consistency.
- Led the design of a unified platform integrating genomic (Gene42), protein (Prot42), and chemical (Chem42) foundation models trained on Cerebras CS-2 hardware.
- Pre-trained and fine-tuned LLaMA2-style models for ultra-long sequence modeling and cross-model interactions in complex R&D environments. Co-authored three foundational model research papers published at NeurIPS 2025.
- Standardized model orchestration by integrating LiteLLM, simplifying multi-LLM switching and unifying API access across the engineering organization.
AI/Machine Learning Engineer
Qritive
- Developed a complete machine learning pipeline, from data acquisition and cleaning to model serving.
- Took three products through the research and development stages to clinical validation.
- Published research papers at the intersection of healthcare and machine learning.
Deep Learning Engineer
Moovita
- Assisted in R&D efforts to develop end-to-end steering control systems for the prototype self-driving system.
- Experimented with recurrent neural networks (LSTM and GRU) to enable automated steering control given the stream of roadview images.
- Prepared a demo to showcase the working of the RNN-based steering control system for the CTO.
Graduate Student Researcher
National University of Singapore
- Researched the latest methods in the domain of self-modifying neural networks.
- Implemented capsule networks for image classification in PyTorch.
- Researched the inner workings of capsule networks and what makes their viewpoint invariant.
- Experimented with a new type of unit inspired by capsule networks that learn time-invariant properties in sequential data (text and audio).
Research Fellow
National Institute of Advanced Studies
- Assisted research efforts in the domain of machine consciousness.
- Implemented the Long Term Recurrent Neural Network(LRCN) architecture to perform image captioning.
- Developed novel methods to determine the level of consciousness of image captioning models.
- Presented research results to the fellowship committee.
Experience
Crowd vs Whale
An AI analyst named CROW runs every 15 minutes across a watchlist of equities. For each ticker, CROW reads recent news from multiple independent sources and searches public institutional filings to assess smart money positioning. It produces a structured psychological analysis, not just a sentiment score, but the specific narrative driving investor positioning, what the crowd appears to be ignoring, and whether institutional data confirms or contradicts that narrative.
When predefined divergence conditions are met, the system fires a named alert with the reasoning pre-written. Every alert is surfaced on the dashboard and published to a dedicated social media account.
Capsule Networks in PyTorch
https://github.com/AahanSingh/Capsule-NetworksNeural Image Captioning
https://github.com/AahanSingh/ConsciousAgentOut-of-domain Object and Style Generation
Education
Master's Degree in Computer Science
National University of Singapore (NUS) - Singapore, Singapore
Bachelor's Degree in Computer Science & Engineering
Ramaiah Institute of Technology - Bangalore
Certifications
Certified Kubernetes Application Developer
Linux Foundation
Skills
Libraries/APIs
PyTorch, TensorFlow
Tools
Helm
Languages
Python 3, Python, Java, C, C++, SQL
Frameworks
LangGraph, Caffe, Agentic Frameworks
Platforms
Linux, Docker, Kubernetes, Web, LangSmith, Langfuse
Storage
Redis, PostgreSQL
Other
Machine Learning, Artificial Intelligence (AI), Deep Learning, Computer Vision, Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Research, Medical Imaging, Software Engineering, Image Classification, Object Detection, Semantic Segmentation, Instance Segmentation, Fine-tuning, Retrieval-augmented Generation (RAG), Big Data, Images, Clinical Validation, Image Processing, Stable Diffusion, DreamBooth, Large Language Models (LLMs), RAG Architecture, RAG Systems, Vector Databases, Machine Learning Operations (MLOps), Data Science, Agentic RAG Systems, RAG Pipelines, Natural Language Processing (NLP), Self-driving Cars, Neural Networks, Operating Systems, Algorithms, Data Structures, Cryptography, IP Networks, Linear Algebra, 3D Graphics, Probability Theory, Graph Theory, Statistics, Statistical Methods, Hypothesis Testing, Generative Adversarial Networks (GANs), Audio, Signal Processing, Health, Dimensionality Reduction, Generative Artificial Intelligence (GenAI), Generative Pre-trained Transformers (GPT), Agentic AI, LangChain, Open-source LLMs, AI Agents, FastAPI, Hugging Face, Langgraph, LiteLLM, Genomics, Multimodal GenAI, Multimodal Models, Cloudflare, Agno, Opik, API Integration, Automated Data Flows
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring