AI & Generative Systems Engineer

Atul Pandey

Buffalo, NY ***** — (1-716-***-**** — ****************@*****.*** — LinkedIn — GitHub — Google Scholar PROFESSIONAL SUMMARY

Artificial Intelligence Engineer building ML, deep-learning, and generative-AI systems from data preparation through monitored production deployment. Strong in Python, PyTorch, RAG, APIs, cloud services, MLOps, model optimization, and translating business problems into measurable AI solutions.

TECHNICAL SKILLS

AI/ML: Python, PyTorch, TensorFlow, Scikit-learn, Transformers, NLP, Computer Vision, Generative AI LLM Systems: LangChain, LlamaIndex, Hugging Face, OpenAI API, RAG, Prompt Engineering, Qdrant/Pinecone Production: FastAPI, REST APIs, Microservices, Docker, Kubernetes, AWS SageMaker, Azure AI, CI/CD MLOps: MLflow, Data/Feature Pipelines, Monitoring, Retraining, Optimization, Statistics, Experiment Evaluation PROFESSIONAL EXPERIENCE

Machine Learning Engineer — Eve Healthcare, New Delhi, India Aug 2023 – Jul 2024

• Designed and trained a PyTorch/MONAI computer-vision pipeline over de-identified medical data, achieving 0.87 AUC and 91% sensitivity across 4 modalities.

• Integrated model inference into FastAPI/Docker services on Azure at sub-800 ms latency, supporting 500+ daily cases with MLflow monitoring.

• Built an XGBoost model and feature pipeline reaching 0.82 AUROC and improving booking conversion 22% across 30+ partner sites. Research Assistant — University at Buffalo, Prof. Junsong Yuan & Sony Corp. Jan 2025 – Sep 2025

• Developed deep-learning pipelines for detection, tracking, and spatio-temporal understanding, achieving 95.37% accuracy and 0.8141 F1.

• Ran 50+ reproducible PyTorch experiments across 3 model families with MLflow tracking, fixed splits, and per-class error analysis.

• Optimized inference 20% through quantization and pruning and delivered tested, version-controlled components through 8 reviews. Assistant Manager, Projects — AdaniConneX, Ahmedabad, India Jun 2021 – Jul 2023

• Built Python/SQL preprocessing and analytics pipelines across 10+ operational KPIs for a 5 MW production environment.

• Automated validated Power BI workflows, eliminating 8+ hours of weekly processing for 12+ stakeholders.

• Translated cross-functional requirements into QA and delivery workflows across 4 teams, reducing rework 20%. PROJECTS

Production Domain RAG & Agent Platform — LangChain, Qdrant, FastAPI Jan 2026 – Present

• Built a RAG platform over 25K synthetic domain documents with hybrid retrieval, reranking, structured prompts, and 6 tool/API integrations.

• Evaluated 500 labelled questions, improving simulated Recall@5 from 0.71 to 0.88 and grounded-answer rate from 76% to 92% while reducing unsupported responses to 6%.

• Deployed containerized inference at 85 requests/second and 210 ms retrieval p95 with monitoring, circuit breakers, and automated regression gates.

Continuous ML Training & Monitoring Pipeline — SageMaker, MLflow, Docker Sep 2025 – Jan 2026

• Automated feature validation, training, tuning, registration, and batch inference across 20+ model configurations and 2M synthetic records.

• Added 35 data, drift, performance, and API tests, detecting 16 seeded regressions and triggering retraining when validation degraded over 2 points.

• Reduced controlled release validation from 68 to 19 minutes while maintaining 98.7% successful pipeline execution across 40 runs. DiffusionPen: Open-Set Handwritten Text Generation — PyTorch Oct 2025 – Present

• Re-engineered a closed-set writer-ID model for open-set generation of custom text from only 4-6 handwriting crops per unseen writer without retraining.

• Diagnosed 3 cross-attention and CFG failures by tracing CANINE-C loading and PAD-mask propagation; restored text-conditioning influence from 0.000 to 0.111 across 2 datasets.

• Scaled PyTorch DDP, AMP, and EMA training across 8 A5000 GPUs for a 14 speedup; benchmarked an EmuruPen hybrid across 3 evaluation dimensions in a reproducible Git pipeline. EDUCATION

University at Buffalo, Buffalo, NY — M.S. Computer Science (Honors) Aug 2024 – May 2026 Coursework: Machine Learning, Deep Learning, Reinforcement Learning Vellore Institute of Technology, India — B.Tech, Electrical & Electronics Engineering Jun 2017 – May 2021 PUBLICATIONS & CERTIFICATIONS

• Publications: Google Scholar: 2 citations, h-index: 1 – Blockchain-AI Integration, IJCRT, 2024; STA-STF RNN for Next Location Prediction, IJFMR, 2024.

• Certification: Deep Learning Specialization, Coursera, 2022.

Contact this candidate