Grounded answers, traceable decisions, and measurable quality.
LOKESH / ENGINEERING PORTFOLIO / 2026
United States · Production GenAI
I build AI systems that earn their way into production.
Production RAG, tool-using agents, evaluation pipelines, and LLM security—designed for accuracy, auditability, and operational control.
100+ engineers supported200+ evaluation cases10× traffic validated
TRACE / REQ-7F3A validated
01Guardrails12ms
02Router18ms
03Retrieval67ms
04Reranker24ms
05Generation302ms
Grounded answer
Response passed citation, policy, and schema validation.
0.94 faithfulness
RAG SYSTEMS AGENTIC WORKFLOWS LLM SECURITY EVALUATION OBSERVABILITY
ABOUT / ENGINEERING APPROACH
I work on the layers that make AI useful after the demo.
My path runs from transformer-based NLP and full-stack AI applications to retrieval systems, bounded agents, evaluation, and production operations. I care about the boundary between model capability and software behavior: what the system knows, what it is allowed to do, how it fails, and how we prove it worked.
Read my engineering philosophyExplicit states, scoped tools, approval gates, and recovery paths.
Security, privacy, observability, and release discipline from day one.
WHAT I BUILD
Reliable intelligence, not demo magic.
The engineering around a model determines whether it can be trusted in production.
Grounded knowledge systems
Hybrid retrieval, reranking, source lineage, and evidence-aware responses.
Bounded agent workflows
Explicit states, scoped tools, approval gates, retries, and verifiable outcomes.
Secure AI applications
Trust boundaries, injection defense, PII handling, output policy, and audit trails.
Evaluation & operations
Task suites, quality regression, latency, token cost, monitoring, and release gates.
INTERACTIVE ARCHITECTURE
Follow a request through the system.
STAGE 03 / PRODUCTION PATH
Retrieval
Hybrid search with filters, query expansion, and lineage.
CONTROLEvidence lineage
FAILURE MODEWeak recall
FEATURED SYSTEMS
Architecture, controls, evaluation, and failure handling.
Retrieval · Security · Evaluation
Enterprise RAG & Security Platform
A production-oriented knowledge system with hybrid retrieval, reranking, evidence lineage, deterministic validation, and defense against malicious retrieved content.
0.94faithfulnessP95latency tracked100%cited claims
Azure OpenAILangChainFastAPIHybrid Search
Agents · Tool Use · Reliability
Agentic Scheduling Platform
A multi-step scheduling workflow that resolves intent, checks policy, invokes bounded tools, verifies outcomes, and escalates safely when confidence is insufficient.
>20%fewer escalations200+test cases10×traffic validation
LangGraphPythonFastAPIAzure OpenAI
RAG · IaC · Developer Experience
Infrastructure Knowledge Assistant
A grounded assistant for infrastructure guidance and generation, pairing curated knowledge with policy-aware templates and validation before output.
100+engineersTracesource lineagePolicyvalidation
PythonFastAPIRAGTerraform
HOW I BUILD PRODUCTION GENAI
Engineering lifecycle
01Use case
02Data classification
03Threat model
04Retrieval
05Evaluation
06Guardrails
07Deployment
08Monitoring
SYNTHETIC EVALUATION
Release candidate 1.8
684msP95 latency1.8kavg tokens0.8%unsupported
Open evaluation lab FAILURE CASE / 03
Vector-only retrieval missed internal language.
Abbreviations and domain phrasing weakened semantic recall. Query expansion, BM25, metadata filters, and reranking produced a stronger evidence set.
BEFOREVector retrievalAFTERHybrid + reranking
Read the decision record CAREER MAP
From applied NLP to production GenAI.
Tata Consultancy Services
01Generative AI Engineer
Engineering production GenAI systems with emphasis on secure retrieval, orchestration, evaluation, and auditability.
Robert Half
02AI Engineer — AI/ML & Agentic Systems
Built an agentic scheduling platform with bounded tool use, test-driven evaluation, observability, and scalable service integration.
BT Group
03Full Stack AI Engineer
Built GenAI-facing infrastructure automation and a RAG knowledge assistant supporting more than 100 engineers.
Tech Mahindra
04Application Development Engineer — AI/ML Practice
Delivered transformer-based document classification and sentiment capabilities for enterprise applications.
TECHNICAL SYSTEM
Skills connected to evidence.
Models & Orchestration
Azure OpenAILangChainLangGraphLlamaIndexStructured outputsRetrieval
Hybrid searchRerankingPineconeFAISSChromaDBWeaviateEvaluation
FaithfulnessAnswer relevancyRetrieval precisionRegression suitesRed teamingPlatform
PythonFastAPIREST APIsSQLDockerAzureAWSSecurity
Prompt injection defensePII redactionTool authorizationAudit loggingPolicy validationOperations
Latency and token metricsTraceabilityVersion trackingFeedback loopsFailure recoveryLET’S BUILD SOMETHING RELIABLE