Headline

AI Engineer | Cut env setup 70% & built 5+ RAG chatbots | AWS Bedrock • Agentic AI • LLMs • Python

About

AI/ML Engineer building production LLM, RAG and Agentic Systems on AWS Bedrock — 5+ chatbots, 10K+ daily queries, 99.9% uptime. Over the past ~3 years, I've shipped production-grade AI systems that go beyond prototypes and actually move metrics: • Agentic AI infrastructure platform — autonomously selects architectures, generates Terraform IaC and orchestrates deployments on AWS. Reduced environment setup from multiple days to ~20 minutes with built-in retry logic (up to 5 attempts) and ghost-resource cleanup. • 5+ production RAG chatbots — using LLMs, vector embeddings and AWS OpenSearch; serving 10K+ daily queries (90-day avg) with ~95% positive feedback/deflection and 99.9% pipeline availability. • Churn prediction with SHAP explainability — 86% recall on ~50K customers (~1M records); SHAP-driven insights informed a 15% cohort churn reduction (3-month post-deployment cohort analysis). • Amazon Nova semi-finalist — Generative AI Advertising Pipeline combining text and image models on AWS Bedrock. • Agent Safety Evaluation Pipeline — benchmarks LLM agent behavior across off-topic rejection, jailbreak attempts and escalation handling using Pydantic for structured, schema-validated scoring. • Prompt-Rubric Co-Optimization Pipeline — iteratively co-evolves prompts and evaluation rubrics to automate alignment between model outputs and task-specific quality criteria. What sets me apart: I have a bias for shipping. When AWS releases a new ML service, I'm testing it the same week. When a library doesn't do what I need, I extend it. I think through multiple approaches before choosing the best one — and then I build it. Tech stack: AWS (Bedrock, OpenSearch, Lambda, SageMaker, ECS/EKS, EC2, S3), Python, LLMs, RAG, Vector Embeddings, Prompt Engineering, AI Safety Evaluation, Pydantic, Docker, Kubernetes, Terraform, CI/CD, Scikit-learn, TensorFlow, PyTorch, SHAP, MLOps, Claude Code, OpenAI Codex, Antigravity. I'm looking for AI/ML Engineer or Generative AI Engineer roles where I can design and deploy cloud-native AI systems that solve hard problems. Open to connect — reach me at anand.071972@gmail.com

Location

Pune District

AI stack

antigravity

claude code

codex

hugging face

ollama

PydanticAI

Work experience

AI Engineer at Mitchell Martin Inc., Apr 2026 - Present · 5 mos, Pune City, Working as a contract AI Engineer embedded with Exact Sciences (acquired by Abbott Healthcare) through Mitchell Martin, focused on production LLM safety and prompt optimization systems. • Designed and implemented an Agent Safety Evaluation Pipeline to systematically benchmark LLM agent behavior across risk categories — off-topic rejection, jailbreak attempts, and escalation handling — using Pydantic for structured, schema-validated scoring outputs. • Developing a Prompt-Rubric Co-Optimization Pipeline to iteratively co-evolve prompts and evaluation rubrics, automating alignment between model outputs and task-specific quality criteria. • Leveraging Claude Code, OpenAI Codex, and Antigravity as AI-assisted development tools to accelerate pipeline design and iteration. Tech stack: Python, LLMs, Pydantic, Prompt Engineering, Claude Code, OpenAI Codex, Antigravity, LangChain, AI Safety Evaluation.

Machine Learning Engineer at Quantiphi, Aug 2023 - Mar 2026 · 2 yrs 8 mos, Mumbai, Design, build and deploy production-grade ML systems on AWS. • Architect and maintain an agentic AI infrastructure platform that automatically selects architectures, generates Terraform IaC and orchestrates deployments — reducing environment setup time by 70% (days to ~20 minutes) with built-in retry logic (up to 5 attempts) and ghost-resource cleanup. • Deployed 5+ production RAG chatbots using LLMs, vector embeddings and AWS OpenSearch; serving 10K+ daily queries (90-day rolling avg) with ~95% positive feedback/deflection rate and 99.9% pipeline availability via CloudWatch Synthetics. • Lead generative-AI pipelines recognised as an Amazon Nova semi-finalist, combining text and image models with AWS Bedrock. • Deploy containerised ML services on AWS ECS/EKS using Docker, Kubernetes and CI/CD; design for high availability and auto-scaling. • Engineered a churn prediction model (86% recall) on ~50K customers (~1M records); SHAP explainability highlighted actionable drivers that informed a 15% cohort churn reduction (3-month post-deployment cohort analysis). • Collaborate with cross-functional teams (product, DevOps, SRE, architects) and mentor junior engineers. Tech stack: AWS (Bedrock, OpenSearch, Lambda, EC2, S3, SageMaker, ECS/EKS), Python, LLMs, RAG, Vector Embeddings, Docker, Kubernetes, Terraform, CI/CD (GitLab), Scikit-learn, TensorFlow, PyTorch, SHAP, MLOps.

Machine Learning Intern at Quantiphi, Jan 2023 - Jul 2023 · 7 mos, Mumbai, Maharashtra, India, Joined as an intern and quickly took ownership of ML workflows and annotation pipelines. • Integrated LLMs and active‑learning techniques into annotation workflows, reducing manual labelling effort by 60%+ and accelerating dataset creation. • Implemented Git‑based version control and CI/CD pipelines across 4+ teams, improving code quality, testing coverage and collaboration. • Built an auto‑annotation pipeline using Python, OpenAI APIs and active learning to triage unlabelled data, speeding up model iterations. • Established engineering best practices for ML workflow optimization, including code reviews, testing and documentation. Tech stack: Python, LLMs, OpenAI APIs, Active Learning, Git, GitLab CI/CD, Docker, Data Labelling Tools.

Software Development Intern at OpTAP Technology Consulting, Jun 2021 - Apr 2022 · 11 mos, India, Developed data‑driven solutions for forecasting and automation in a consulting environment. • Designed and implemented time‑series forecasting models (ARIMA, Prophet) that increased prediction accuracy by 30%, informing strategic inventory planning. • Built a sales forecasting system using Python and Pandas; communicated insights to stakeholders and refined models based on feedback. • Automated data‑entry workflows with OCR and Selenium, eliminating 100% of manual data entry and saving 20+ hours/week for client reporting. • Developed modular scripts and deployed them as micro‑services to integrate with client systems. Tech stack: Python, Pandas, NumPy, Statsmodels, Prophet, Selenium, Tesseract OCR, OpenCV, Git.

Education

Bachelor of Engineering - BE, Computer Science, CHANDIGARH UNIVERSITY

HSC (12th), Non-Medical, KV NAD Karanja