Saketh Bollina

AI/ML Engineer @ CareCentrix

About

I'm an AI/ML Engineer with 4 years of experience. I build production-grade AI systems that move real business numbers, not just model metrics. What sets my work apart is where I start. Every engagement begins with the business problem, the KPI to move, and the ROI case before a single model is chosen. That discipline comes from pairing deep technical work in LLM fine-tuning, RAG architecture, agentic AI orchestration, and MLOps with the mindset of a trained business strategist. I've worked directly with C-suite, legal, finance, and operations leaders to scope AI around P&L impact, translate technical findings into executive-ready ROI narratives, and build the explainability and governance frameworks regulated enterprises need. I'm currently seeking roles where I can keep building AI products that reduce cost, accelerate revenue-driving decisions, and earn genuine business adoption. Always open to connecting with people working at the intersection of AI and business strategy. Outside of work, I'm happiest learning something new. I was a national-level swimmer, I play golf, and I'm currently teaching myself to surf, which is the latest in a long habit of picking up new hobbies just to see how far I can take them. That same curiosity shows up in how I build, including plenty of vibe coding on the side.

Country

City

United States

Industry

Higher Education

Skill

Artificial Intelligence (AI), Generative AI, AI & ML Foundation, Cloud Certification, Foundation Models, pytest, REST APIs, Pandas (Software), OpenAI API, BERT (Language Model), Microsoft Power BI, XGBoost, LightGBM, Microsoft Azure, Amazon Web Services (AWS), TruLens, Scikit-Learn, Airflow, DVCProHD, hugging face transformers

Experience

AI/ML Engineer

CareCentrix

2025-9 - Present · 1 yr 1 mo

Connecticut, United States

• Fine-tuned BERT, GPT-4, and LLaMA with QLoRA for contract classification, obligation extraction, and summarization, cutting analyst review time 75% and improving accuracy 28% across thousands of monthly contracts. • Built production RAG pipelines with LangChain and vector databases (FAISS, Pinecone) grounded in the company's policy library, adding RAGAS retrieval scoring that gave compliance teams an auditable quality signal and cut hallucinations 40%. • Designed multi-agent LangGraph orchestration with risk-tiered escalation and human-in-the-loop approval gates, removing 60% of manual touchpoints and freeing analysts for higher-value work. • Presented platform ROI to leadership: per-document cost down 35%, turnaround from days to under four hours, and 3x throughput with no added headcount. • Deployed MLOps on AWS SageMaker with MLflow, DVC, and Kubernetes at 99.9% uptime across 10,000+ requests, adding SHAP/LIME explainability and drift monitoring that cut silent model failures 45%.

AI Engineer

Blend360 Communications

2023-6 - 2024-7 · 1 yr 2 mos

Hyderabad, India

• Led discovery workshops with C-suite and VP-level leaders at Fortune 1000 clients, turning vague mandates into KPI-aligned ML problem definitions. Drove a 22% accuracy gain and 30% cut in operational costs per engagement. • Deployed a fine-tuned BERT sentiment and intent system for a financial services client handling thousands of daily support interactions, surfacing escalation risk 3x faster than legacy reporting and reducing high-value account churn. • Architected a RAG-based editorial research system (LangChain, OpenAI API, Pinecone) for a media client, using TruLens to validate retrieval quality. Cut research time per article 55% with no added headcount. • Delivered a collaborative filtering and LLM-augmented recommendation engine for a retail client, validated via A/B test. Drove a 25% lift in click-through and 18% drop in churn.

Python Developer

Aricent

2021-6 - 2023-5 · 2 yrs

Hyderabad, India

• Built Python backend services and automation frameworks for carrier-grade telecom software across 100+ networks serving 500M+ end-users, where defects carried SLA penalties and emergency patch costs. Made reliability and proactive defect prevention core priorities. • Applied Scikit-learn classification and anomaly detection to real-time network telemetry, prioritizing high-risk test scenarios before release. Cut critical defect escape rates 30%, sparing carrier clients costly emergency patches and SLA penalties. • Engineered NLP pipelines to extract, normalize, and classify high-volume unstructured network logs and fault records, building the tokenization, entity extraction, and feature engineering foundation that now informs my LLM and RAG work. • Designed reusable REST API libraries and CI/CD-integrated regression pipelines shared across product teams. Accelerated feature delivery 25%, lowered post-release defects, and set a modular architecture adopted as the team standard.