Home

ACHYUTHA PRANAVI K - Gen AI Engineer

[email protected]

Location: Jersey City, New Jersey, USA

Relocation: YES

Visa: H1B

Resume file: ACHYUTHA PRANAVI K_Gen_AI_Engineer_1769549325107.pdf Please check the file(s) for viruses. Files are checked manually and then made available for download.

ACHYUTHA PRANAVI K [email protected] || +1 201 430 5523 || LinkedIn PROFESSIONAL SUMMARY Delivered production-grade Generative AI systems and LLM-powered assistants across semiconductor, automotive, manufacturing, and enterprise engineering domains, enabling documentation intelligence, knowledge retrieval, and decision-support workflows. Built and scaled enterprise RAG platforms using Azure OpenAI (GPT-4), LangChain, and LangGraph, providing grounded, citation-backed responses over large technical corpora under strict latency and cost constraints. Implemented multi-agent GenAI workflows (planner, retriever, validator, compliance) with LangGraph, automating ingestion, summarization, reasoning, and validation pipelines and improving workflow throughput by ~35% with auditability. Fine-tuned and optimized open-source LLMs (LLaMA, Mistral) using PEFT (LoRA / QLoRA) for domain-specific summarization, classification, and reasoning, achieving ~15% accuracy gains with reduced compute overhead. Engineered prompt engineering and instruction-tuning strategies, including dynamic prompt routing and temperature control, to balance response quality, safety, and inference cost across enterprise GenAI workloads. Built production-grade RAG systems integrating Azure Cognitive Search (BM25 + vector) with FAISS, Pinecone, and ChromaDB, reducing hallucinations by ~30% while improving retrieval precision and traceability. Developed secure, scalable GenAI inference services using FastAPI and RESTful APIs with OAuth2/RBAC, structured logging, request tracing, and streaming responses for real-time and batch workloads. Containerized and deployed GenAI and ML services using Docker and CI/CD pipelines, with exposure to Kubernetes-based deployments, enabling reliable scaling and controlled production rollouts. Designed and implemented ML and GenAI pipelines spanning data ingestion, feature engineering, model development, evaluation, and deployment using PySpark and cloud-native data platforms. Established LLM evaluation and benchmarking frameworks combining automated metrics (ROUGE, BLEU, Retrieval@k) and human-in-the-loop review to assess accuracy, latency, grounding quality, and cost. Implemented Responsible AI and safety controls including content moderation, prompt guardrails, explainability (SHAP, LIME), and audit logging to support compliant enterprise adoption. Built multi-turn conversational memory and session-level state management for GenAI assistants, improving dialogue continuity, contextual recall, and reducing redundant interactions. Integrated GenAI systems into enterprise applications and internal platforms, enabling AI-powered copilots, auto-drafting workflows, and intelligent search aligned with business processes. Engineered scalable data ingestion and enrichment pipelines using PySpark and cloud data services to process structured telemetry, logs, and unstructured documents for downstream ML and GenAI use cases. Designed observability and monitoring dashboards exposing GenAI usage, latency, cost, reliability, and error patterns, reducing MTTR by ~30% and improving operational visibility. Collaborated closely with engineering, data science, platform, and operations teams to productionize GenAI solutions aligned with security, compliance, and real-world deployment constraints. Progressively transitioned from predictive analytics and ML into hands-on delivery of Generative AI systems, contributing across design, optimization, evaluation, deployment, and reliability. Demonstrated consistent delivery of scalable, business-aligned AI systems that balance accuracy, performance, cost efficiency, safety, and long-term maintainability. TECHNICAL SKILLS Languages Python, SQL, JavaScript, Bash, R Generative AI & LLMs GPT-4, Azure OpenAI, OpenAI APIs, LLaMA, Mistral, Claude, Cohere, prompt engineering, few-shot prompting, instruction tuning, structured prompts, chain-of-thought prompting, LLM orchestration, prompt routing, safety-aware generation Agentic AI & Orchestration LangChain, LangGraph, LlamaIndex, multi-agent workflows, planner-retriever-validator patterns, tool calling, stateful workflows, multi-step reasoning RAG & Vector Retrieval Azure Cognitive Search (BM25 + vector), FAISS, Pinecone, ChromaDB, hybrid retrieval, chunking strategies, semantic similarity, re-ranking, citation grounding, context filtering, retrieval optimization LLM Fine-Tuning & Optimization PEFT, LoRA, QLoRA, task-specific fine-tuning, instruction-tuned models, evaluation-driven model selection Machine Learning & Deep Learning PyTorch, TensorFlow, Keras, scikit-learn, XGBoost, LightGBM, CatBoost, supervised learning, ensemble models NLP & Text Analytics BERT, Hugging Face Transformers, spaCy, named entity recognition, document classification, summarization, sentiment analysis, text normalization, preprocessing pipelines Time-Series & Forecasting ARIMA, Prophet, LSTM, demand forecasting, trend analysis, anomaly detection MLOps, APIs & Deployment FastAPI, RESTful APIs, Docker, CI/CD pipelines, MLflow, model versioning, batch inference, near-real-time inference, A/B testing, basic drift detection, Kubernetes (AKS) Cloud & Data Engineering Microsoft Azure, AWS, Azure Databricks, Azure Data Factory, AWS S3, AWS Lambda, EC2, Redshift, Glue, Athena, EMR, PySpark, Spark, Delta Lake, Airflow, Kafka Databases & Data Stores PostgreSQL, MySQL, Azure Synapse, Amazon Redshift, SQL Server, MongoDB, Cassandra Document AI & Multimodal Systems Azure Vision, OCR pipelines, OpenCV, document intelligence workflows, structured data extraction Monitoring, Evaluation & Responsible AI Azure Monitor, Application Insights, ROUGE, BLEU, retrieval@k, human-in-the-loop evaluation, SHAP, LIME, content moderation, prompt guardrails, responsible AI practices PROFESSIONAL EXPERIENCE Gen AI engineer May 2024-Present Intel, New York Enterprise Generative AI Platforms Silicon, Manufacturing & Internal Engineering Systems Responsibilities: Designed and implemented LLM-powered engineering assistants using Azure OpenAI (GPT-4), LangChain, and LangGraph to support silicon design documentation search, manufacturing knowledge retrieval, and internal engineering workflows, reducing manual lookup and review effort. Implemented multi-step, agent-based GenAI workflows with LangGraph to automate ingestion, summarization, validation, and reasoning over technical specifications, manufacturing logs, and internal wikis, improving workflow throughput by ~35%. Built production-grade RAG pipelines using Azure Cognitive Search (BM25 + vector), FAISS, ChromaDB, and Pinecone to query hardware specifications, firmware documentation, APIs, and SOPs, reducing hallucinations by ~30% and improving citation accuracy. Applied query-time context filtering, document chunking strategies, semantic similarity scoring, and lightweight re-ranking to improve retrieval precision across large technical corpora while controlling token usage and inference latency. Fine-tuned open-source LLMs (LLaMA, Mistral) using PyTorch PEFT (LoRA / QLoRA) for technical summarization, component classification, and log interpretation, achieving ~15% accuracy improvement on internal evaluation benchmarks. Implemented prompt routing and model-selection logic to route complex or safety-sensitive queries to GPT-4, while serving routine engineering queries with optimized open-source models, reducing overall inference cost by ~20%. Developed LLM-powered engineering copilots using LangChain and Azure OpenAI to assist with firmware development, driver APIs, system configuration, and internal knowledge discovery, improving onboarding efficiency and developer productivity. Built secure GenAI inference APIs using FastAPI with OAuth2/RBAC, structured logging, request tracing, and streaming responses to integrate LLM capabilities into internal engineering platforms. Containerized GenAI services using Docker with exposure to Azure Kubernetes Service (AKS), enabling scalable batch and near real-time inference with safe rollout patterns such as blue-green and canary releases. Integrated vector stores and metadata layers using Azure Cognitive Search and Delta Lake to support traceability, reproducibility, and controlled access to AI-assisted engineering workflows. Engineered data ingestion and enrichment pipelines using PySpark, pandas, and Azure Data Factory to process structured telemetry, system logs, and unstructured technical documents for downstream GenAI applications. Implemented Responsible AI controls using Azure AI Content Safety, prompt guardrails, system constraints, and explainability techniques to support safe, reliable, and policy-compliant AI behavior. Designed LLM evaluation and benchmarking workflows combining automated metrics and human-in-the-loop review to measure response accuracy, latency, cost, and grounding quality across GPT-4 and open-source models. Developed multi-turn conversational memory with session-level state management to support long-running engineering troubleshooting and diagnostic conversations. Built observability and feedback tooling exposing retrieved context, citations, confidence indicators, and latency metrics, enabling rapid iteration and continuous improvement of GenAI systems. Delivered Power BI dashboards visualizing GenAI adoption, usage patterns, latency, cost, and reliability metrics for technical leads and engineering stakeholders. Integrated monitoring and alerting using Azure Monitor and Application Insights, reducing mean time to resolution (MTTR) by ~30% and maintaining high service reliability. Worked closely with senior engineers and architects on model selection, retrieval strategies, safety controls, and deployment patterns for enterprise GenAI systems. Collaborated with silicon engineering, manufacturing, platform, data science, and IT teams to productionize GenAI capabilities and support enterprise-wide adoption. Environment: GPT-4, LLaMA, Mistral, LangChain, LangGraph, Azure OpenAI, Azure Cognitive Search, FAISS, Pinecone, ChromaDB, FastAPI, Docker, Azure Kubernetes Service (AKS), PyTorch (LoRA / QLoRA), PySpark, Delta Lake, Azure Data Factory, Power BI, Azure Monitor, Application Insights AI/ML Engineer / Data Scientist Nov 2021 Apr 2024 General Motors., Detroit, MI Fraud Detection, Warranty Analytics & Enterprise Risk Platforms Responsibilities: Designed and deployed supervised fraud detection and risk-scoring systems using Python, XGBoost, Random Forest, and Azure ML to identify anomalous warranty claims and dealer fraud patterns, reducing false positives by ~18% and improving investigator throughput. Engineered ensemble learning approaches combining gradient boosting and tree-based models to stabilize fraud predictions across heterogeneous dealer, vehicle, and supplier datasets, improving score consistency under evolving data distributions. Built large-scale risk stratification pipelines on Azure Databricks using PySpark to continuously rank high-risk vehicles, dealers, and claims, outperforming legacy rule-based detection approaches in early-intervention accuracy. Developed feature engineering frameworks integrating vehicle telemetry aggregates, warranty histories, repair frequencies, dealer behavior signals, parts usage patterns, and geographic risk indicators, improving downstream model discrimination power. Applied transformer-based NLP models using BERT, Hugging Face Transformers, and PyTorch to analyze technician notes, warranty descriptions, and customer complaints, improving text classification accuracy by ~20% for investigative workflows. Designed scalable NLP processing pipelines with spaCy and Azure AI Language to perform entity extraction, root-cause categorization, sentiment analysis, and text normalization over unstructured service documentation. Implemented document intelligence and OCR workflows using Azure Vision APIs to extract structured data from scanned invoices, warranty forms, and inspection reports, significantly reducing manual review effort. Architected secure, versioned ETL pipelines using Azure Data Factory and Delta Lake to ingest, validate, and lineage-track warranty, dealer, and financial datasets supporting enterprise risk analytics. Leveraged Azure AutoML and MLflow to accelerate experimentation, hyperparameter tuning, and lifecycle tracking of fraud and risk models while maintaining reproducibility and audit readiness. Applied explainable AI techniques using SHAP and LIME to surface feature contributions and model rationale for high-impact fraud decisions, increasing investigator trust and supporting audit and compliance reviews. In later phases prototyped LLM-assisted audit and investigation workflows using LangChain and Azure OpenAI to summarize claim histories, surface policy references, and aggregate supporting evidence, reducing manual investigation effort by ~35%. Designed batch and near real-time inference workflows to integrate ML risk scores into downstream warranty management, dealer oversight, and enterprise risk platforms with low operational latency. Containerized and deployed ML inference services using Docker, with exposure to Azure Kubernetes Service (AKS) and CI/CD pipelines in Azure DevOps, supporting reliable production rollouts and controlled rollback strategies. Implemented model performance monitoring and basic drift detection using Azure Monitor and scheduled evaluation jobs to ensure long-term stability and reliability of fraud and risk models. Developed executive and operational dashboards using Power BI to visualize fraud risk scores, model performance trends, dealer behavior patterns, and operational KPIs for leadership and business stakeholders. Collaborated closely with warranty operations, finance, data engineering, and compliance teams to productionize ML systems aligned with GM s enterprise risk controls, regulatory requirements, and real-world operational constraints. Environment: Python, SQL, PySpark, XGBoost, Random Forest, PyTorch, TensorFlow, BERT, Hugging Face Transformers, spaCy, LangChain, Azure OpenAI, Azure ML, Azure Databricks, Delta Lake, Azure Data Factory, Azure Vision, Azure AI Language, MLflow, Docker, Azure Kubernetes Service (AKS), Azure DevOps, Power BI, SHAP, LIME AI/ML Engineer Jan 2019 Dec 2020 Smartous LLC, India Enterprise Retail Analytics & Demand Forecasting Platforms Responsibilities: Designed ML-ready data pipelines using Python, PySpark, and AWS (S3, Glue, Redshift) to support enterprise demand forecasting, pricing optimization, and inventory planning across thousands of SKUs and multiple fulfillment regions. Engineered feature engineering frameworks with SQL, dbt, and Snowflake to generate time-series, customer-level, and product-level features, improving downstream model stability across seasonal and high-volume product categories. Developed and deployed demand forecasting systems using ARIMA, Prophet, LSTM, and neural networks to model seasonality, trends, and promotional effects, improving forecast accuracy by ~25% for high-impact retail segments. Implemented model evaluation and backtesting pipelines using RMSE, MAPE, and rolling-window validation to compare statistical and deep-learning models prior to production rollout, reducing forecast error volatility. Built ensemble forecasting approaches combining statistical methods and neural models to improve robustness under demand spikes and promotional volatility, supporting more reliable inventory and replenishment decisions. Integrated external business and market signals including promotions, pricing changes, demand trends, and calendar effects as model features, improving sensitivity to short-term demand shifts and promotional accuracy. Operationalized forecasting models through batch and scheduled inference pipelines using AWS Lambda and CI/CD workflows, enabling reliable execution aligned with enterprise planning cadences. Designed analytics and decision-support dashboards using Power BI, Tableau, and Amazon QuickSight to surface forecast outputs, confidence intervals, and demand risk indicators for merchandising and supply-chain teams. Conducted A/B testing, uplift analysis, and cohort analysis to quantify the business impact of model-driven pricing and promotion strategies, informing data-backed rollout decisions. Implemented data and model governance practices using AWS Lake Formation, versioned datasets, and access controls to ensure reproducibility, auditability, and secure ML workflows. Collaborated with DevOps teams to productionize ML pipelines with AWS CloudWatch monitoring and alerting, improving pipeline reliability and operational visibility. Partnered with merchandising, supply-chain, and e-commerce teams to translate ML forecasts into actionable inventory, pricing, and replenishment decisions, increasing adoption of data-driven planning. Environment: Python, SQL, PySpark, ARIMA, Prophet, LSTM, Neural Networks, AWS (S3, Glue, Redshift, Lambda, Athena, Lake Formation, CloudWatch), Snowflake, dbt, Power BI, Tableau, Amazon QuickSight, Git, CI/CD ML Engineer Sep 2017 Dec 2018 Cisco Systems, Chennai India Predictive Analytics & Enterprise Risk Intelligence Responsibilities: Designed supervised predictive analytics systems using Python, scikit-learn, pandas, Random Forest, and Gradient Boosting to support customer churn prediction, contract risk analysis, and anomaly detection across large enterprise customer portfolios. Built and evaluated classification and ensemble models including logistic regression, tree-based methods, and boosting techniques to identify high-risk customer accounts and abnormal usage patterns, reducing noise in downstream analytics and renewal workflows. Developed feature engineering pipelines using SQL and Python to construct customer-, contract-, usage-, and product-level features from CRM platforms, subscription systems, billing data, and network telemetry summaries. Implemented customer segmentation and risk stratification analyses using clustering algorithms and statistical techniques to surface churn drivers and usage behavior patterns, enabling more targeted renewal and retention strategies. Contributed to automated ETL workflows using AWS Glue, Python, and SQL to ingest, cleanse, and transform data from CRM, licensing, billing, and third-party systems into analytics-ready datasets. Supported batch model training and scoring pipelines using AWS (S3, Redshift, Athena, Lambda, EC2) to enable scalable risk scoring, periodic churn assessment, and enterprise reporting in a secure cloud environment. Optimized analytical SQL queries in Amazon Redshift to improve performance of recurring churn and risk reports, reducing query latency and improving dashboard responsiveness. Applied statistical modeling techniques using SAS (PROC LOGISTIC, PROC REG) alongside Python-based ML models to support pricing analysis, renewal likelihood estimation, and customer behavior insights. Built and maintained executive and operational dashboards using Tableau and Power BI to track churn trends, risk indicators, customer adoption metrics, and portfolio-level KPIs for analytics and business stakeholders. Participated in model validation and monitoring activities, tracking accuracy, stability metrics, and early drift indicators to support reliable production analytics and reporting outputs. Supported enterprise data governance and compliance practices by contributing documentation, data lineage, and control artifacts aligned with Cisco s internal risk, security, and analytics governance frameworks. Collaborated with data engineers, business analysts, and DevOps teams to integrate model outputs into reporting layers, renewal planning workflows, and decision-support systems. Gained hands-on experience across the full ML lifecycle, including data preparation, feature engineering, model development, evaluation, basic deployment, and monitoring within large-scale enterprise environments. Environment: Python, SQL, SAS, scikit-learn, pandas, NumPy, Random Forest, Gradient Boosting, AWS (S3, Redshift, Athena, Glue, Lambda, EC2), Tableau, Power BI, ETL Pipelines Python Developer/Data Analyst Feb 2016 - Aug 2017 Square Panda India Pvt Ltd Telecom Analytics & Customer Intelligence Responsibilities: Analyzed large-scale telecom usage and operational datasets using Python (pandas, NumPy) and SQL to support network performance monitoring, cost optimization, and customer intelligence across prepaid and postpaid segments. Designed analytical data models and reporting datasets integrating call-detail records (CDRs), service logs, and customer profiles, improving consistency and reliability of downstream analytics used by operations and business teams. Built customer behavior and churn analysis pipelines using Python and R, identifying usage drop-offs, service-quality drivers, and lifecycle risk patterns to inform targeted retention strategies. Developed predictive churn-risk and call-drop models using logistic regression and decision trees, enabling proactive customer outreach and data-driven network optimization for high-risk cohorts. Engineered automated ETL workflows using Python, SQL Server, MySQL, and Oracle to ingest, cleanse, and standardize telecom data from multiple source systems, reducing manual preparation effort and reporting delays. Implemented data validation, reconciliation, and anomaly-detection checks within ETL pipelines to improve data accuracy, completeness, and trustworthiness for analytics and reporting workflows. Performed market segmentation and cohort analysis based on usage patterns, demographics, and service plans, supporting pricing optimization and customer acquisition initiatives. Prepared labeled datasets and baseline features for early churn prediction and network-quality models, supporting data science teams during experimentation and validation phases. Designed and executed A/B testing frameworks to evaluate service changes, feature rollouts, and customer-experience improvements, enabling statistically driven product and operational decisions. Conducted VoIP and call-quality analytics to identify network bottlenecks, dropped-call patterns, and service degradation issues, contributing insights for capacity planning and infrastructure optimization. Supported database migration and consolidation initiatives by validating data integrity, schema mappings, and historical consistency across legacy and modernized data platforms. Collaborated with engineering, operations, and marketing teams to translate analytical insights into actionable improvements across customer experience, network operations, and retention programs. Environment: Python, R, SQL (SQL Server, MySQL, Oracle), pandas, NumPy, Logistic Regression, Decision Trees, Tableau, Power BI, ETL Pipelines EDUCATION Master s in computer science - University of Central Missouri | January 2021 December 2022 Bachelor s in computer science - Lovely Professional University | August 2011 May 2015 CERTIFICATIONS Microsoft Certified: Azure AI Engineer Associate - 2023 AWS Certified Machine Learning - Specialty 2021 Keywords: continuous integration continuous deployment artificial intelligence machine learning business intelligence sthree rlang information technology Michigan