Hey, I'm Abhishek GuptaBackend & AI Engineer
I help SaaS founders and agencies build reliable backend systems, APIs, AI integrations, RAG solutions, automation workflows, and cloud-ready products. Available for remote projects and global collaborations.
Work Experience
Engineered complex Agentic RAG pipelines to optimize context retrieval for LLMs, improving response relevance and accuracy.
Developed a custom BM25 retrieval system and integrated it into a hybrid search engine to enhance document ranking with Re-ranking.
Accelerated development cycles by 30% through the strategic integration of LLM-based coding assistants and automation tools.
Education
Skills
Languages
Frameworks & Libraries
Databases & AI Tools
Cloud & MLOps
Projects
Mushroom Classification Model
Processed a dataset of 3 million samples, optimizing feature engineering to increase model performance by 2%. Trained ensemble models achieving 97.5% test accuracy by implementing rigorous cross-validation to prevent overfitting.
View ProjectUser Chat-Style Mimicking with Qwen3-8B
Fine-tuned Qwen3-8B on custom Hinglish WhatsApp and synthetic datasets to replicate specific linguistic nuances. Implemented QLoRA for parameter-efficient fine-tuning, reducing memory overhead and cutting training time by over 50%. Evaluated model output using an LLM-as-a-judge framework, achieving a 5% increase in stylistic accuracy compared to baseline few-shot prompting.
View ProjectEnd to End Credit Card Fraud Detection System
Orchestrated a robust MLOps pipeline using Apache Airflow for automated data ingestion and Scikit-learn for predictive modeling. Deployed a real-time inference layer via FastAPI and Docker Compose on AWS EC2, ensuring high availability and low latency. Managed data and model lineage using DVC and MLflow (DagsHub) to ensure 100% experiment reproducibility. Configured a monitoring stack with Prometheus and Grafana to track drift and system health in a production environment. Automated the CI/CD lifecycle using GitHub Actions, pushing containerized images to Amazon ECR for secure management.
View Project
End-to-End AI Refund Agent
Engineered an autonomous financial LLM agent utilizing the OpenAI Agents SDK armed with programmatic tools to parse complex customer claims and execute secure transactions. Implemented a high-precision RAG architecture via Pinecone Hybrid Search and structured output schemas alongside explicit input/output guardrails to guarantee deterministic, fraud-resistant execution. Configured complete observability loops using OpenInference and Langfuse for real-time execution tracing and telemetry, monitoring cost metrics and agent tool-calling accuracy.
View ProjectAchievements
IEEE International Hackathon
Finalist
Smart India hackathon
Finalist
KRIYETA 2.0
2nd Runner Up