ECMS REQ ID

542983

PU

CMTADM

Client Name

AT&T; Services Inc..

Number of Openings

4

Country

India

Detailed JD (Roles and Responsibilities)

Role Summary: Own agentic workflows (LangGraph), MCP tools, retrieval/embedding pipelines, and model evaluation. Translate ambiguous problems into robust, measurable solutions with explicit documentation and business impact.

Key Responsibilities

· Build agentic workflows in LangGraph; create reusable templates for multi-tool agents.

· Design, implement, and operate MCP servers/tools to expose APIs, data access, and actions.

· Apply advanced prompt engineering; maintain a versioned prompt registry with telemetry and A/B tests.

· Build embedding pipelines for semantic search/classification/clustering/retrieval; integrate with downstream apps (intent detection, topic modeling, deduplication, ranking).

· Apply dimensionality reduction and similarity search (PCA/t-SNE/UMAP; cosine/Euclidean; FAISS/ScaNN).

· Build and evaluate ML models (regression, random forest, XGBoost/LightGBM, SVM, Naive Bayes, k-means,



hierarchical clustering) with sound diagnostics and inference.

· Design experiments (A/B/MVT/DOE) and causal analyses (PSM, causal forests, DiD); translate into actionable insights.

· Partner with Full Stack and DevOps on data contracts, latency/SLOs, observability, and deployment.

Required Technical Skills:

· Python (advanced): FastAPI, async I/O, packaging, pytest; Linux proficiency

· GenAI/Agentic: LangGraph (templates/orchestration), MCP servers/tools, LLM tool-use/function calling, retrieval, streaming

· LLM Fine-tuning & Eval: SFT/ORPO/DPO; prompt/response evaluation frameworks; guardrails

· Embeddings & Vector DBs: TF-IDF, Word2Vec, GloVe, FastText, BERT/SBERT, OpenAI/Azure OpenAI; FAISS, Azure AI Search, Pinecone, ScaNN

· Statistics & Causal: hypothesis testing, CIs, bootstrapping, Bayesian basics; feature selection (Lasso/Ridge, RFE); SHAP

· Data Eng basics: PySpark, SQL; Azure Databricks, Data Factor

📌 AT&T -Gen AI (India)
🏢 Clifyx
📍 India