I started my career as a data engineer — building ETL pipelines, wrangling messy financial data at scale, and making sure numbers landed where they needed to be. Somewhere along the way, I realized I wasn't just moving data around. I was looking for patterns, automating decisions, and trying to make systems smarter.
That's what pulled me into AI. Not the hype, but the realization that the best automation doesn't just move data — it understands it.
Today I build production AI systems — RAG pipelines, agentic workflows, LLM-powered features — but the foundation is still the same: clean data in, intelligent decisions out. Whether it was processing 3TB+ daily financial data at LTIMindtree or building a privacy-safe RAG chatbot as a solo AI engineer at Rebecca Everlene, the thread has always been: find the mess, build the system, make it disappear.
I think of engineering the way I think of Sudoku — every constraint is a clue, every bottleneck is a puzzle, and the satisfaction isn't in the answer, it's in the path you took to get there.
Currently building with LangGraph, MCP, RAGAS, Claude Code, and Cursor — shipping to real users, not just prototyping.
📍 Open to US-based roles | 📧 anusreemohanan22@gmail.com
|
AI & Agents
|
Cloud & Data
|
|
Languages
|
Analytics & Dashboards
|
|
Tools
|
|
|
RAG-powered medical assistant processing 27K+ documents at 94% accuracy with sub-500ms response time. Disease risk prediction with Random Forest & XGBoost. OCR document analysis via Claude Code.
|
Voice-to-CRM pipeline transcribing 500+ daily sales calls via Google Cloud Speech-to-Text, extracting structured CRM fields with Gemini 2.5 Pro at 92% accuracy. Zero manual entry.
|
|
Fine-tuned Microsoft BioGPT for medical QA using LoRA (PEFT), reducing perplexity from ~1.9M → 29.3. Ray Tune hyperparameter optimization proved the approach after naive fine-tuning degraded performance.
|
Multi-agent automation in n8n with controller-based routing across WriterAgent, IdeaAgent, and SummaryAgent. Automated email delivery with fallback routing. 100% reduction in manual effort.
|
|
RAG-based career platform matching resumes against 50K+ job postings and generating personalized learning paths across 100K+ resources. Built on Snowflake Cortex with dbt medallion architecture.
|
Multi-agent inventory optimizer using DQN & PPO reinforcement learning. 98.85% service level with 39% cost reduction over heuristic baselines. Custom agentic tools for simulation and decision explanation.
|
📂 More Projects
| Project | Description | Tech |
|---|---|---|
| UrbanCrash Analytics | Traffic collision data platform processing 2.4M+ records, Power BI & Tableau dashboards | ER Studio, Talend, Python |
| Telecom Churn Prediction | Churn model at 85% accuracy with SHAP explainability, reducing churn by 18% | H2O AutoML, SHAP |
📄 Stock Market Prediction Using ML Techniques
Published in IJSART, Volume 6, Issue 1 — Jan 2020
🔗 Read the paper