New Delhi, Delhi, India
I build reliable end-to-end GenAI systems, my expertise lies in fine-tuning LLMs/SLMs to evaluation and scalable deployment. My superpower is turning POCs into production-ready solutions that are fast, measurable, and secure. What I Bring: • LLM Training & Fine-Tuning: LoRA/QLoRA + DeepSpeed/FSDP for domain models (3B–40B). • Inference at Scale: vLLM/TGI + monitoring to support real workloads and latency targets. • GenAI Product Delivery: For code generation, NL→SQL, RAG, and agentic workflows. • Applied ML: forecasting, anomaly detection, and computer vision across enterprise use cases. Keywords for Recruiters: #GenAI #LLM #FineTuning #AIOps #Python #AWS #Azure #FastAPI #RAG #MCP #AgenticAI #CI_CD #ComputerVision Let’s Connect: If you’re building GenAI products, happy to chat.
* Trained and fine-tuned LLMs (Falcon, CodeLlama, LLaMA2, Pythia, MPT, Flan‑T5) on NVIDIA A100 clusters. * Delivered 20+ POCs and 8 MVPs across GenAI/NLP/CV and moved multiple solutions into production. * Pre-trained and instruction-tuned LLMs (3B–40B) using FSDP/DeepSpeed + PEFT (LoRA/QLoRA); built a copilot exceeding GPT‑3.5 on internal HumanEval.
• Extensive experience in custom Language Model (LLM) training, fine-tuning like Pythia, mpt, open-assist, starcoder, codegen, flan-T5, gpt4all, dolly, camel, vicuna. • LLM deployment for large-scale projects, successfully accommodating and serving a user base exceeding 5000+ on an AWS EC2 with 4 A10 GPUs. • Demonstrated proficiency by successfully developing more than 18 proof of concepts (POCs) and 8 Minimum Viable Products (MVPs) in diverse domains such as computer vision, natural language processing (NLP), and predictive analytics. • Led a team of 6 machine learning engineers, and provided expert guidance and mentorship to 10 interns in the field of AI/ML. • Seasoned machine learning consultant with a strong track record of assessing business requirements and presenting viable machine learning solutions to clients and prospects in South East Asia, the US, and Canada regions.
- As part of the Center of Excellence team, designing and building solutions for various business use cases around Cognitive and Predictive Analytics. - Interacting with clients and prospects to analyze their business requirements and present feasible solutions.
Codezoned is an Open source non-profit-organization which is targeting on educating and spreading our core belief of "learning by doing". >> Responsibilities undertook by me - > Mentored students on Algorithms and Data Structures. > Built various repositories to escalate open-source culture in colleges. > Actively contributed to open-source projects like Psipher and ScriptsDump (100+ stars on Git Hub). > Guided 15+ students for open-source, personal projects and research areas like Hyperspectral and Image Processing.
> The thesis internship was conducted under the guidance of Sh. T S Rawat Sc - ‘F’ at DTRL, DRDO. • Review of ensemble-based decision making criteria for the classification of Hyperspectral Images. • Investigated supervised and unsupervised band selection methods for Hyperspectral Images. • Developed an Unsupervised Band Selection model based on Evolutionary Multiobjective Optimization for Hyperspectral Images using Boltzmann Entropy. This resulted in an increase in overall classification accuracy by 10%. • Evaluated various entropies for Hyperspectral Images to find the balance between the number of features/bands and usefulness of the information provided by the mathematical function.
> Interned at Defence Terrain Research Laboratory (DTRL). • Designed and Developed an algorithm for an automatic spectra verification in terms of their quality and correctness using the proposed parameters by the lab. • Techniques like Empirical Mode Decomposition and Spectral Similarity Measure were used to find a degree of similarity between the various bands of the spectrum.