Agents that reason, retrieve, verify, and recover
GenAI / Agentic AI Engineer
My path into agentic AI is grounded in building systems around the model, not treating a prompt as the product. Across two in-progress agent workflows, I am applying typed state, hybrid retrieval, bounded self-correction, source-aware evaluation, and human review so generated answers remain inspectable and useful.
Why this direction fits
I bring a backend-first approach to agentic systems: explicit contracts, controlled tool access, observable execution, and evaluation against defined targets.
Current evidence
BCA in Information Technology, IIT Dharwad-aligned NSQF L5.0 training, Oracle Data Science Professional, and an in-progress multi-agent research system.
01 / CAPABILITY MAP
How I would contribute
Orchestration
Stateful LangGraph workflows with planner, retriever, fact-checker, and correction loops.
Grounding
Hybrid retrieval with Qdrant or Chroma, BM25, schema context, and structured fallbacks.
Guardrails
Pydantic contracts, PII redaction, prompt-injection checks, read-only tools, and bounded retries.
Evaluation
Faithfulness, relevancy, context recall, tracing, and target-based regression checks.
02 / VERIFIED SKILLS
Role-aligned toolkit
- Python 3.13
- LangGraph
- LangChain v1
- FastAPI
- Qdrant
- Chroma
- BM25
- Pydantic v2
- Langfuse
- Docker
03 / SELECTED WORK
Evidence for this role
Enterprise Multi-Agent Knowledge & Research Assistant
A cyclical Planner, Retriever, and Fact-Checker workflow combining hybrid and vectorless RAG with enterprise guardrails.
- Target: 34% lower context hallucination
- Evaluate: faithfulness · relevancy · recall
Self-Healing BI & SQL Analytics Agent
A read-only analytics agent that inspects schemas, generates and validates SQL, self-corrects failures, and creates reviewed insights.
- Target: self-correction within three retries
- Safety: read-only execution · human review
High-Throughput Asynchronous AI Gateway & Task Engine
An asynchronous gateway with distributed tasks, provider fallback, token-bucket limits, deduplication, and auditable usage.
- Target: 250+ concurrent requests
- Target: sub-50ms queue latency · >80% coverage
04 / TAILORED RESUME
Resume for this role
GenAI / Agentic AI Engineer
This version will foreground the experience and project evidence most relevant to this hiring path.
- —Agent orchestration and self-correcting workflows
- —Retrieval, grounding, and output validation
- —LLM evaluation, safety, and observability