Bhavana Sunkari

AI/ML • Middle • 27 years • Chicago, USA • UTC-06:00 • Relocation • Work permits: United States • Available in 2 weeks • from $80k per year
Software Developer
6+ years exp
6+ years ML exp
SQL
JavaScript
TypeScript
C
C++
Python
Java
Active 14 days ago
+1 (312) 6849688
Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles

Overview

A practical LLM-focused engineer at a solid early-middle level who implements RAG-style document Q&A and LLM observability plumbing. The strongest proven skill is LLM systems engineering and observability, demonstrated by the instrumented tracked_chat wrapper and Prometheus metrics (llm-cost-monitor/app/wrapper.py and app/metrics.py). There is little to no evidence of custom model training, formal evaluation pipelines, unit tests, or production-grade deployment automation in public code.

Technical skills

SQL
JavaScript
TypeScript
C
C++• Middle • 6y+
Python• Middle • 6y+
Java• Middle • 6y+
C++
TensorFlow C++
Python
Flask
FastAPI• 6y+
Java
Spring Boot• 6y+
Databases
DynamoDB
FAISS
MySQL
Pinecone
PostgreSQL
Redis
AI/ML
AI Agents
LLM
Prompt Engineering
RAG
Semantic Search
Google GenAI SDK
Groq
huggingface_hub
Scikit-learn• 6y+
TensorFlow• 6y+
LangChain
LangGraph
NLP
DevOps
Amazon EC2
Amazon EKS
AWS Lambda
Rest API
Vector
Prometheus
CI/CD• 6y+
Git• 6y+
Jenkins• 6y+
AWS
Azure
Docker
GitHub Actions
Kubernetes
Frontend
GraphQL
QA
Postman

Timeline

Software Development Engineer – AI/ML & Backend Systems • Middle
UnitedHealth Group • Full-Time
Oct 2024 to Present 1 Year 10 Months Chicago In office
Developed an AI knowledge retrieval platform using RAG with semantic search, serving enterprise users and reducing manual search effort. Designed agentic automation workflows using LangChain and LangGraph to improve response accuracy and speed for enterprise queries. Built and scaled asynchronous FastAPI microservices and optimized PostgreSQL and caching to maintain high availability and throughput. Implemented CI/CD with GitHub Actions plus Docker and Kubernetes, and added monitoring/observability for reliable production operations.
FastAPI
LangChain
LangGraph
Redis
PostgreSQL
GitHub Actions
Docker
Kubernetes
CI/CD
AWS
Azure
Illinois Institute of Technology (Illinois Tech)
Master's Degree • Computer Science
2023–2025 Chicago, Illinois
Software Development Engineer • Middle
Bosch • Full-Time
Apr 2020 to Jul 2023 3 Years 3 Months In office
Built distributed backend systems and REST APIs for large-scale industrial IoT event processing. Developed predictive maintenance solutions using machine learning models and telemetry/sensor analytics to reduce downtime. Improved integration latency by redesigning API and integration services with Spring Boot and FastAPI. Automated testing and deployment with Jenkins and Git, and refactored monoliths into modular service architectures while applying NLP retrieval techniques using TensorFlow and scikit-learn.
Python
Java
C++
Spring Boot
FastAPIsince 2020
Jenkins
Git
TensorFlow
Scikit-learn
CI/CDsince 2020
Middle AI/ML Engineer Confidence: Medium LLM Engineer
A practical LLM-focused engineer at a solid early-middle level who implements RAG-style document Q&A and LLM observability plumbing. The strongest proven skill is LLM systems engineering and observability, demonstrated by the instrumented tracked_chat wrapper and Prometheus metrics (llm-cost-monitor/app/wrapper.py and app/metrics.py). There is little to no evidence of custom model training, formal evaluation pipelines, unit tests, or production-grade deployment automation in public code.
Model Architecture & Training
1/10
How well models are designed and trained
Minimal model-building or training work; evidence is limited to inference-time choices (model selection, generation config) rather than architecture or training.
Evidence
RAG-based-AI-Chatbot/app.py:get_model_name (model discovery) and ask_gemini (client.models.generate_content use and GenerateContentConfig)
Data Pipeline & Feature Engineering
3/10
How data is prepared for models
Clear, practical document ingestion and chunking pipeline for PDF RAG use-cases using LangChain loaders and a text splitter; solid engineering for indexing but no advanced feature engineering.
Evidence
RAG-based-AI-Chatbot/app.py:process_files (PyPDFLoader, RecursiveCharacterTextSplitter, chunk creation)
RAG-based-AI-Chatbot/app.py:process_files (preserving source_file metadata on pages)
Experimentation & Evaluation
1/10
How results are measured and tested
Little-to-no experimentation or evaluation infrastructure for models or ablations; basic runtime prompts used only to exercise the monitor.
Evidence
llm-cost-monitor/app/main.py (sample PROMPTS loop used for live telemetry rather than formal experiments)
llm-cost-monitor/app/metrics.py (defines Prometheus metrics but not experiment/eval orchestration)
MLOps & Deployment
3/10
How models are shipped to production
Practical MLOps/observability work for LLM usage - Prometheus metrics, an HTTP metrics endpoint, and an instrumented wrapper showing awareness of monitoring and cost accounting.
Evidence
llm-cost-monitor/app/metrics.py (REQUEST_LATENCY, TOKENS_USED, COST_TOTAL, REQUEST_COUNT and start_metrics_server)
llm-cost-monitor/app/wrapper.py:tracked_chat (instrumentation of tokens, cost, request latency and request counters)
Computational Efficiency
1/10
How efficiently computing resources are used
Minimal computational-efficiency engineering; no evidence of GPU optimization, batching strategies for throughput, quantization, or profiling.
Evidence
llm-cost-monitor/app/wrapper.py (token-based cost accounting indicates cost-awareness but not runtime efficiency engineering)
Research Depth & Innovation
1/10
Depth of research and new ideas
No research-level or novel algorithmic work; the code integrates existing libraries and shows pragmatic engineering rather than research innovation.
Evidence
RAG-based-AI-Chatbot/app.py (standard RAG pattern using LangChain loaders and Chroma; fallback/retry logic is pragmatic but not research)
Expertise
RAG• Middle
Conversational AI & Chatbots• Middle
MLOps & Model Lifecycle• Middle
Industries
Artificial Intelligence• Middle
Technologies
Python• Middle • 6y+
SQL
C++• Middle • 6y+
MySQL
PostgreSQL
Redis
LangGraph
Rest API
LangChain
Flask
Pinecone
DynamoDB
FAISS
Groq
GitHub Actions
FastAPI• 6y+
Prometheus
Scikit-learn• 6y+
Prompt Engineering
AI Agents
NLP
Google GenAI SDK
Azure
huggingface_hub
CI/CD• 6y+
Jenkins• 6y+
TensorFlow• 6y+
Git• 6y+
AWS
Docker
Kubernetes
LLM
RAG
TensorFlow C++
Amazon EKS
AWS Lambda
Amazon EC2
Vector
Semantic Search
Recommendations
  • Develop production-grade CI tests and automated unit/integration tests for the ingestion and wrapper components (e.g., tests for process_files and tracked_chat).
  • Add a reproducible evaluation pipeline and experiment tracking (W&B or MLflow) for measuring retrieval+generation quality and latency trade-offs.
  • Harden production readiness: add input sanitization, stronger error classification, logging/structured traces, and a deployment recipe (Dockerfile + k8s/argo) with SLI/SLO considerations.
  • Improve cost/efficiency work by adding batching strategies, token-level profiling, and optional model-usage throttling or circuit-breaking based on metrics.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
RAG-based-AI-Chatbot llm-cost-monitor