Abnik Ahilasamy

AI / ML Engineer

I am an AI / ML engineer with an M.S. in Data Science from Arizona State University. I specialize in agentic AI systems, retrieval-augmented generation, and LLM fine-tuning, with hands-on experience building production-ready pipelines that emphasize reliability, latency, and real-world impact.

Work Experience

CIPS-AI Labs Post-Training Researcher Jun 2026 - Present
  • Built modular LLM and agent evaluation pipelines for retrieval, reasoning, memory, and tools.
  • Developed dataset curation, synthetic tests, schema validation, and model evaluation.
  • Added OpenTelemetry tracing and tuned request controls to improve evaluation reliability.
Larsen & Toubro AI / LLM Engineer Apr 2024 - Jun 2024
  • Built a 0 to 1 agentic RAG system, cutting hallucinations 35%.
  • Owned retrieval, reasoning, APIs, evaluation, and production infrastructure.
  • Cut latency 40% with caching, batching, and memory optimization.
Excelerate Data Analyst Aug 2023 - Sep 2023
  • Built Power BI dashboards with Power Query and DAX for real-time KPI tracking.
  • Cleaned and analyzed business data in Excel, improving decision efficiency by 30%.
  • Delivered stakeholder reports that reduced analysis turnaround time by 25%.
LetsGrowMore ML Engineer Jun 2023 - Jul 2023
  • Built ML pipelines across NLP, time-series, and vision using Python and Scikit-learn.
  • Developed LSTM, RNN, and CNN models for classification and emotion detection.

Education

Arizona State University Masters

M.S. in Data Science

GPA 3.97/4.0. Coursework includes Statistical Machine Learning, Data Processing at Scale, Computer Vision, and Data Mining.

SRM Institute of Science and Technology Bachelors

B.Tech. in AI & ML

Top 5%, GPA 3.8/4. Coursework includes Deep Learning, Artificial Intelligence, Neural Networks, NLP, and Algorithms.

Featured Projects

View all
RAG SQL LangGraph

Aker RAG

Property-scoped AI platform with deterministic SQL, guarded LLM-to-SQL, Qdrant retrieval, and markdown/UI response blocks.

GitHub Read more
Agents Safety Python

AffectKit

Emotion-state middleware for LLM agents with profiles, mood decay, emotional memory, and safety-bounded prompt modulation.

GitHub Read more
GitHub Jira Slack

GitHub Jira Sync

Webhook backend that comments on Jira issues from PRs and sends Slack alerts when GitHub and Jira states drift.

GitHub Read more
Attention Neural Nets VLM

From Scratch

Personal reference repo for foundational AI models, including attention implementations, neural networks, and VLM fine-tuning notes.

GitHub Read more

Blogs / Writing

All posts
Can You Trick LLMs Into Using Fewer Tokens? The Metrics That Actually Matter in AI Agents What Building an AI System Taught Me About Reliability Why You Should Hire Me - From an AI

Things I enjoy

  • I like watching technical talks, AI demos, startup, product videos, and engineering deep-dives.
  • I follow cricket, F1, business podcasts, and long-form technical blogs, and I also love sharing what I learn on LinkedIn and X.