● AI Focus

Gajendra Saraswat

AI / ML Engineer • Python • PyTorch • RAG • Agent Design
3+ years blending enterprise backend with hands-on ML — fraud, medical imaging, compression. Daily Windsurf/Devin with RAG/Agent governance (BNY AI Pioneer L3). Springer publication, 500+ LeetCode.
AT A GLANCE
AI: PyTorch, TensorFlow, Scikit-learn, RAG, Agents
Impact: 97.6% fraud • ~90% leukemia • 24.2× compression
Prod: Flask/Docker/Heroku + Spring/Kafka
Honors: Springer • GATE 99.1% • Kaggle 19%
kaggle.com/sarques • github.com/saarques
TECHNICAL SKILLS — AI LENS
Languages: Python, Java (8,17), C++, SQL
Data Science & AI: Scikit-learn, TensorFlow, PyTorch, Pandas, Numpy, OpenCV
LLM & AI Eng: RAG, Agent Design, Prompt Eng, Nuggets, Windsurf, Devin
ML Eng: Model dev/deploy, Feature eng, SMOTE, Precision/Recall
Backend for ML: Spring Boot, REST, Microservices, Kafka, Flask
MLOps: Docker, Jenkins, GitLab CI/CD, GAE, Redis, Vertica, DB2
Quality: Grafana, Splunk, Orion, PVT, JUnit, Mockito, MCP/Jira
Security: RBAC, VMAD, SiteMinder, AI governance
EXPERIENCE — AI + BACKEND INTEGRATION

BNY Mellon — Software Engineer | Aug 2023 – Present

Chennai • Blending backend + ML

Backend & ML Integration (Contract Platform)

Enhanced Spring Boot microservices with validation/anomaly detection; 36% faster via parallelization/batching. Secure REST/Kafka via API Gateway. Deployed via GAE/Docker/GitLab CI/CD — strong fundamentals for scaling ML systems to production.

Data-Driven Quality & AI Tooling

Python metrics/dashboards + proactive anomaly detection. NextGen 300+/95% → fully automated MCP/Jira regression. Daily Windsurf/Devin (RAG/Agent) for code-gen/debugging — applying AI Pioneer L3 (RAG/Nuggets/orchestration) to backend & ML workflows.

System Architecture & Responsible AI

DR for 25+ apps multi-cluster, Kafka backup failover (8 topics, 2 clusters). Audited Devin CoT secret exposure, rotated secrets, authored governance guidelines — AI risk & compliance in regulated finance.

SELECTED PROJECTS — AI
Autoencoder-Huffman Ensemble — Dec 2024 • Springer link →

Two-stage encoder-decoder + Huffman. MNIST/CIFAR-10/Fashion MNIST with minimal loss. Achieved 5.5×–24.2× compression (15.6 MNIST, 24.2 CIFAR-10 without overhead). Published Springer Singapore.

PyTorch • Autoencoders • Huffman • Python
Credit Card Fraud Detection — GitHub 16★/16 forks →

Logistic Regression (balanced weights) 97.6% acc, 89.6% recall • SMOTE for imbalance • Flask REST, validation, Docker/Heroku. Final Year Project.

Python • Scikit-learn • Flask • SMOTE • Docker
Leukemia Detection (PyTorch/SeResNext) — GitHub →

Automated leukemia classification on microscopic cell images addressing morphological similarity. ~90% accuracy via hyperparameter tuning, Pandas/NumPy preprocessing, Docker Hub.

PyTorch • Keras • SeResNext • Docker
EDUCATION & PUBLICATION
NIT Warangal — M.Tech CS, CGPA 7.13 (Sep 2021–Jun 2023)
MBM Jodhpur — B.E. CS, CGPA 7.48 (Aug 2016–Oct 2020)
Springer Dec 2024: Autoencoder-Huffman Ensemble Model for Image Compression — 5.5×–24.2× link →
Kaggle: SIIM-ISIC Melanoma Top 19% • 2× Expert
AI Pioneer L3: RAG, Agent Design, Prompt Eng, Nuggets, Workflow Orchestration
Download Resume
One-pager & detailed versions.
AI One-Pager PDF Detailed (2-page) PDF
Key strengths:
• Imbalanced data (SMOTE) & eval
• Compression & vision pipelines
• RAG/Agent + MLOps (Docker/Heroku)
• Backend for ML (Kafka/REST)
CERTIFICATIONS
AI Pioneer L3 — BNY (RAG/Agents)
Machine Learning — Coursera
Deep Learning Specialization
Math for ML — Coursera
CONTACT
saraswatgajendra97@gmail.com
+91 9784537496
linkedin.com/in/sarques