Skip to content
View yash176A's full-sized avatar
🎯
Focusing
🎯
Focusing
1 follower · 0 following

Block or report yash176A

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
yash176A/README.md

Yashwanth Sena Veerupakshi, AI and ML Engineer

LinkedIn Email Profile views

⏳ 4+ years
ML in production		 📦 5 repos
tested, benchmarked, documented		 🏆 3 certifications
LangChain · Python DS · Power BI		 🎓 MS in IT
University of Cincinnati, 2025

🧭 The arc

Started at Amazon as a Data Scientist doing classic ML at scale: forecasting, ETL over millions of daily records, XGBoost in production through SageMaker. The pivot point came with the shift to generative AI: the interesting failures moved from "the model is inaccurate" to "the model is confidently wrong", and that pulled me into LLM systems, retrieval, and evaluation. Now I build RAG architectures, multi-agent workflows, and the monitoring and evaluation infrastructure that keeps them honest in production.

🔭 Current focus

  • Enterprise LLM systems for document intelligence and semantic search
  • RAG architectures balancing retrieval depth, latency, and hallucination rates
  • Multi-agent orchestration with task decomposition and self-correction loops
  • Evaluation frameworks for grounding accuracy and LLM observability

🧱 Building blocks

Core tools

Domain Stack Depth
LLMs and RAG LangGraph, LangChain, FAISS, Pinecone, ChromaDB, Hugging Face ██████████ daily driver
ML and DL PyTorch, TensorFlow, scikit-learn, XGBoost, LightGBM ██████████ daily driver
MLOps MLflow, drift detection, CI/CD for ML, model versioning ████████░░ production experience
Serving FastAPI, Docker, AWS SageMaker, GCP AI Platform ████████░░ production experience
Data ETL pipelines, SQL, pandas, large-scale preprocessing ████████░░ production experience
Analytics Power BI, Tableau, Plotly, Seaborn ██████░░░░ working proficiency

📌 The proof: five deep-dive repos

Every number in these READMEs was measured, every behavior claimed is tested, and every design decision has an ADR.

Repo What it shows The one detail worth clicking for
drift-sentinel Statistical drift detection as a FastAPI service The test suite caught a real KS-test false positive; the fix is a commit you can read
rag-evalkit Offline-first RAG evaluation with CI quality gates Zero-dependency core; its own CI gates on mrr=0.9 using the tool itself
agentic-extract Self-correcting multi-agent extraction on LangGraph Validator errors literally become the next attempt's prompt feedback, provably, in tests
attention-lab Transformer from scratch in PyTorch Causality proven by gradient, parity with nn.MultiheadAttention to 1e-5
tabular-ml-pipeline Gated ETL, tuned XGBoost, MLflow tracking Training physically cannot accept data that failed the quality gate

🤝 Quick connect

LinkedIn   Email

The fastest way to evaluate me is to open any repo above and read one ADR.

Pinned Loading

  1. demand-forecasting-lstm demand-forecasting-lstm Public

    Multivariate retail demand forecasting with a PyTorch LSTM. Direct 7-day-ahead forecasts from price, promotion, weather, and calendar signals, benchmarked against seasonal-naive and linear baseline…

    Python

  2. pricing-engine pricing-engine Public

    Dynamic pricing engine: an XGBoost demand model plus a profit-maximizing optimizer, served via FastAPI with a versioned model registry. Evaluated honestly against a known demand function, capturing…

    Python

  3. Selfcorrecting-rag Selfcorrecting-rag Public

    Self-correcting RAG pipeline: a LangGraph evaluator loop that grades every answer for grounding and relevance, refines weak retrievals, and abstains instead of hallucinating. Runs offline with zero…

    Python

  4. agentic-extract agentic-extract Public

    Self-correcting multi-agent document extraction on LangGraph. Validator feedback loops into retry attempts, bounded escalation to human review, pluggable LLM backends.

    Python

  5. drift-sentinel drift-sentinel Public

    Statistical drift detection service for production ML. PSI, KS and Jensen-Shannon detectors behind a FastAPI scoring API, with Docker, CI and reproducible benchmarks.

    Python

  6. tabular-ml-pipeline tabular-ml-pipeline Public

    Tabular ML pipeline where training cannot run on ungated data: enforced quality gates with per-defect audit counts, tuned XGBoost, MLflow tracking.

    Python