Software Engineer specializing in AI with 5 years of hands-on experience developing scalable systems for LLM finetuning, generative AI, RAG pipelines, and cloud-native backends.
At Humanitarian AI, I helped build a no-code platform to fine-tune open-source LLMs from Hugging Face. I worked on pipelines to fine-tune models like Llama 3 and DeepSeek using Unsloth PEFT (LoRA/QLoRA), automated dataset processing with Airflow and Snowflake, and implemented a RAG module using Gretel, Pinecone, and Neo4j. I also designed secure deployments with MLflow + AWS SageMaker and built backend services, integrating Lambda, API Gateway, and Cognito for access and metadata management.
Engineered a context-aware AI agent chatbot using LangChain and PaLM-2 LLM on GCP Vertex AI, integrating advanced reasoning, memory, and tool orchestration to enable dynamic interactions and boost user engagement by 40%. Improved retrieval accuracy by 60% with a robust RAG pipeline using Pinecone Vector DB, and implemented evaluation/monitoring with Ragas and LangSmith for automated debugging and iterative improvements. Built a scalable backend with FastAPI, a React/TypeScript frontend, and secure inference using Guardrails API, while accelerating deployment by 40% through enhanced CI/CD workflows with Docker and Cloud Build. Leveraged Looker for data mining and insights on customer behavior.
Engineered and scaled high-availability Postgres RDS clusters powering a distributed 3D optimization platform with 100k+ daily users.
Improved system performance by 25% through schema design, indexing, and advanced query tuning while supporting replication, failover, and DR workflows.
Led production on-call, migrations, upgrades, and security hardening, ensuring reliability, compliance, and seamless business continuity.
