Headline

Principal GenAI Architect & Applied AI Researcher | Scaling & optimizing LLMs/SLMs — efficient adaptation (LoRA/QLoRA, distillation), RAG, agentic systems, distributed training (FSDP/vLLM) | AI at scale

About

Most people in AI have ridden one wave. I've built through four — large-scale software and Big Data, deep learning, computer vision on the edge, and today's Generative AI. The throughline has never changed: turn hard, ambiguous problems into systems that understand, scale, and last. Today I'm a principal-level Generative AI architect and applied researcher with 16+ years designing, scaling, and shipping AI for enterprise and regulated domains. My focus is GenAI that holds up in production — LLMs and SLMs, retrieval-augmented generation, agentic workflows, and the efficient adaptation and serving that make them fast, affordable, and reliable at scale. That focus stands on real foundations. I've led deep-learning and computer-vision teams running real-time analytics on edge devices, architected large-scale data platforms processing terabytes, and built the backend systems beneath consumer products used by millions. Each era compounded the last — which gives me a rare, full-stack view of how AI actually gets built, scaled, and trusted, not just prototyped. What I bring is range with depth: the engineering to work the hard layers — efficient model adaptation, distributed training, retrieval, and evaluation; the judgment to ship into high-stakes domains with safety, privacy, cost, and reliability as first-class constraints; and the leadership to build and mentor cross-functional teams (data science, MLOps, backend, embedded) and own roadmaps and architecture standards. I turn ambiguity into measurable, dependable outcomes. I stay close to frontier GenAI research and believe the best systems are both inventive and reliable. If you're building or scaling GenAI — and want it leaner, faster, and genuinely production-ready — let's connect.

Location

Bengaluru

AI stack

anthropic apis

claude code

claude

cursor

deepseek

gemini

Work experience

ML Advisor at Cigna Healthcare, Oct 2023 - Present · 3 yrs 1 mo, Bengaluru, Karnataka, India, Lead the Generative AI charter within Cigna's AI Center of Excellence (AICoE) — owning architecture, delivery, and engineering standards for enterprise GenAI across insurance and healthcare workflows. - My edge is efficiency at scale: adapting and serving LLMs and SLMs so they hit real latency, cost, and reliability targets. I work the hard layers — parameter-efficient adaptation (LoRA/QLoRA, knowledge distillation, quantization, pruning), retrieval and grounding (RAG, vector search, cross-modal), and GPU-scale training and inference (PyTorch, DDP/FSDP, ZeRO-style sharding, FlashAttention, vLLM-style serving). - Designed and shipped Project Genius, a reusable RAG platform for prompt-driven and Q&A use cases — turning ambiguous business problems into production-ready solutions and onboarding new workflows with minimal rework, accelerating go-to-market. - XMRetriever — cross-modal retrieval that cut prompt tokens ~75% and reduced hallucination from 8.6% to 1.2%. - Currently building agentic systems: a decentralized agent swarm (HIVE) and a literature-grounded AutoML engine that auto-selects the optimal model and adaptation strategy per task (SAGE). - Drove model strategy end-to-end — prompt engineering, few-shot, CoT/ToT reasoning, and targeted LLM/SLM adaptation (LoRA/QLoRA) — for strong domain grounding and accuracy. - Established LLM evaluation gates and guardrails — PII and data-integrity controls, prompt-injection defenses, and exact-match plus hallucination scoring — gating every release for security and compliance in regulated domains. - Set production serving patterns (vLLM-style) that balanced latency, throughput, and cost against clear SLOs. - Led technical execution across Product, Security, and Data Engineering; ran architecture reviews and mentored engineers on eval-first development.

NLP Architect at Hexad GmbH, Feb 2022 - May 2023 · 1 yr 4 mos, Wolfsburg, Lower Saxony, Germany, Information extraction from pdfs through computer vision task and impose NLP processing for relevant information extraction. Worked on pruning of neural networks without compromising of much accuracy drop in terms of FLOPs and inferencing speeds. Worked on PoCs related to LLMs. Worked on question answering modules using BERT/GPT 3.5 with RAG for retrieval from vector databases. Extensive experience with prominent libraries and frameworks including Huggingface (BERT, GPT-3, ColBert), Bi-Encoders, Cross-Encoders, and Haystack. My proficiency extends to deploying Knowledge-Intensive NLP solutions and leveraging Weak Supervision techniques. Led the successful deployment of state-of-the-art models in industry-leading companies, contributing to improved user experiences and increased efficiency.Developed end-to-end AI/ML solutions that integrated seamlessly with existing systems, resulting in streamlined workflows and reduced operational complexities.Pioneered the implementation of advanced NLP techniques, enhancing language understanding and information retrieval capabilities for complex datasets.

Practice Lead - R&D at Trigyn Technologies, Sep 2020 - Feb 2022 · 1 yr 6 mos, Bengaluru, Karnataka, India, Guiding the evolution of video analytics in R&D using advanced techniques such as computer vision, deep learning, model quantization, and data curation. Spearheading the creation of a state-of-the-art video analytics solution, meticulously engineered for availability, scalability, and reliability. Orchestrating the infusion of intelligent insights into surveillance camera feeds, seamlessly integrating with expansive Command & Control infrastructure on a grand scale. Distinguished by a holistic approach that not only adapts traditional video analytics paradigms but also deftly avoids invasive middleware components. Pioneering the optimization of video analytics computation on existing infrastructure with minimal disruption. Cultivating strategic partnerships with product and engineering units, while harmonizing efforts across diverse teams. Championing strategies that empower clients to harness analytics for a data-centric culture shift. Collaborating with operations to achieve optimal project delivery efficiency, ensuring timely deployment of high-impact, data-driven insights. Committed to propelling security enhancements and operational streamlining through ingenious technology breakthroughs in every facet of video analytics.

Head of AI - R&D at Inkers Technology, Jan 2019 - Sep 2020 · 1 yr 9 mos, Bengaluru, Karnataka, India, Leads development of analytical models using statistical, machine learning and data mining models. Defines model development tactics. Work with senior management / Director(s) across product, client, taking a large role in driving that agenda with business units. Provide expertise on mathematical concepts and inspire the adoption of advanced analytics and data science across the breath of the organisation. Closely interacting with client to understand the business needs and proposing effective and optimised solution. Establish partnerships with product and engineering teams and work closely with other teams. Develop strategies to help clients activate their analytics and create a more data-driven culture. Work with team and operations to maximize project delivery quality, timeliness and efficiencies in order to deliver high- value and relevant insights to our business partners as quickly and effectively as possible

Scientist at Inkers Technology, May 2016 - Sep 2020 · 4 yrs 5 mos, Bangalore, Working on Image processing using Deep Learning, NeuralNets, CNN, RNN. Analysing image and infer object detection, bounding boxes, captioning , activity detection (temporal data), finding depth and normals using VGG, ResNet, Mobilenet, EfficientNet (B0-5). Specialising in training Deep Learning model on computer vision and also visualising neural network attentions to see how model is doing. It includes Grad-CAM, neurons activations, layers visualisation etc. I have trained several models on object detection (large as satellite-images and small as cashews nuts), gaze detection, facial attributes (like gender, emotion, age, pose on single model). Summary of my DL/ML proficiency skills are: PyTorch (training on clusters, optimising neural networks, Grad-CAM, statistical modeling, 2D, 3D, convolution, depth convolution, SK-Net optimised, trained with depth information for Medical science (CT-scan, MRI etc)) Python (Threading and multiprocessing; Flask; scikit-learn, bayesian modelling, matplotlib, ggplot, visualising techniques) Shell scripting (running jobs from kernel, shell scripting, AWK, ruby, well versed with regex for fast parsing while making training datasets, cronjob initiation) Data Analysis (extensive analysis on data before going for training, visualisation in all respects to check the data forming which normalisation, plotting, features extraction and normalising decision factor) Deployment (deployment of product using docker-image with pip requirements file, Jira, cluster Jupiter notebook management, kubernetes scalable. Optimization (quantisation, model running by hyper parameters, prunning network layers, int8,int16,FP16 conversion while training) Model testing (test benchmark on trained model to check how good is model, Normalise Mutual Information, Adjust Mutual Information, R-precision, ROC/AUC curve, Precision, Recall, specificity (confusion matrix), F1-score) Squeezing model if size and accuracy is a BIG concern to run on embedded device

Big Data Developer at Impetus Infotech Pvt. Ltd., Jan 2012 - Jan 2016 · 4 yrs 1 mo, Bangalore, Analysing large data-sets to provide strategic direction to the company. Performed quantitive analysis to make prediction accuracy. Developed statistical model to enhance the business growth of the company. Familiarity with data mining, machine learning, artificial intelligence, applied mathematics, statistical inference and modelling. Ability to manipulate data-sets. Solving complex problem for large data-set using Big Data technologies. Desire to analyze large data sets, find the truth in data, and develop efficient processes for data analysis. Search query optimisation on large data-sets using elastic search and solr. Competence in building predictive models and validating these models against hypothesis. Team player and self-starter with effective communication, motivation and organizational skills combined with attention to detail and business process improvements.

Education

Master of Technology - MTech, Artificial Intelligence and Machine Learning, Birla Institute of Technology and Science, Pilani

Bachelor's degree, Computer Science, Visvesvaraya Technological University

Public links

github: https://github.com/nsinghtelligence

navneet.page: https://www.navneet.page/