Headline

Senior AI/ML Engineer (Serving Notice Period) specialising in Generative AI solutions | RAG | Agentic AI | LLMOps | Fine-Tuning

About

Senior AI/ML Engineer with 5 years 8 months building production Generative AI and Agentic AI systems for Apple Inc. and CVS Healthcare (Aetna) — at Infosys. I design and ship systems that are production ready and scalable. RAG pipelines with hybrid retrieval (BM25 + semantic, RRF re-ranking), multi-agent orchestration with LangChain/LangGraph/MCP, fine-tuned open-weight LLMs (Llama-3.2-3B via LoRA/QLoRA on Unsloth), and Azure-native LLMOps (MLflow, LangSmith, Docker, Kubernetes, CI/CD). What I've delivered: → Architected and built an enterprise-grade multi-agent SRE Copilot platform (LangGraph orchestration, Mem0 persistent memory, MCP-based tool integration into Grafana/Loki/Mimir/Tempo/Splunk, Agentic RAG over Confluence, NeMo Guardrails for AI safety, LangFuse/ClickHouse observability) enabling conversational incident diagnosis across 5+ enterprise observability platforms at CVS Healthcare/Aetna. → Architected a 3-tiered LLM allocation + data-egress-prevention layer that cut LLM operational costs 40-50% while scaling from 3 to 150+ applications at CVS Healthcare/Aetna. → Built a codebase-vectorization RAG pipeline (FAISS, MongoDB Atlas Vector, Azure AI Search) that cut LLM token consumption ~60% across 15+ repos at Apple's Emerging Technologies & Services org — adopted by 30+ engineers. → Shipped LLM-powered test generation tooling (LangChain + custom MCP servers) handling 500+ daily specification requests, cutting manual authoring time ~70% across 20+ modules. → Built failure-classification ML models (Scikit-learn) hitting ~88% accuracy across 10+ failure categories, reducing mean-time-to-triage ~35%. Core stack: Python · LLMs · RAG · Agentic AI (ReAct, planning, router-specialist, tool-calling patterns) · LangChain · LangGraph · MCP · Fine-Tuning (LoRA/QLoRA/PEFT) · HuggingFace · PyTorch · FAISS · Azure AI Search · pgvector · Azure ML · Azure OpenAI · Docker · Kubernetes · MLflow · LangSmith · FastAPI · Responsible AI & Guardrails. Recognized 3x as Platinum Talent at Infosys, with Outstanding performance ratings in 4 of the last 5 years. Currently exploring Senior/Staff AI-ML Engineer and Generative AI Engineer roles where I can take production LLM systems from prototype to scale. Open to connecting with fellow GenAI builders, recruiters, and hiring managers.

Location

Pune City

AI stack

antigravity

codex

Work experience

Senior AI/ML Engineer (Specialist Programmer L2) — Generative AI & LLM Systems at Infosys, Jun 2023 - Present · 3 yrs 3 mos, Pune District, Delivering Agentic AI systems for CVS Healthcare’s Aetna Insights Observability Platform. SRE COPILOT — MULTI-AGENT AI PLATFORM FOR INCIDENT DIAGNOSIS & OBSERVABILITY • Architected a multi-agent SRE Copilot using LangGraph to orchestrate specialized agents for conversational incident diagnosis and failure investigation. • Implemented persistent memory (Mem0) for cross-session context retention; built Redis-based prompt caching to reduce latency. • Integrated MCP-based tool connectivity to Loki, Grafana, Mimir, Tempo, and Splunk for real-time telemetry retrieval within agent workflows. • Built an Agentic RAG architecture over Confluence for context-aware operational guidance, served via Azure AI Foundry/Azure OpenAI. • Built observability pipelines (LangFuse, ClickHouse, Redis) for prompt tracing, token analytics, and audit logging. • Designed AI safety guardrails (NVIDIA NeMo Guardrails) against prompt injection, jailbreaks, and data exfiltration; ran adversarial red-teaming. • Built evaluation pipelines for agent/RAG faithfulness, groundedness, and tool-selection accuracy. Impact: unified incident diagnosis across 5+ observability platforms with a production multi-agent architecture, reducing MTTR. LLM COST OPTIMIZATION & DATA EGRESS PREVENTION • Architected a 3-tiered LLM allocation pattern (self-hosted Qwen3-Coder:30B, GPT-3.5-turbo, gated GPT-4-class endpoints) behind a zero-trust perimeter — cut LLM costs 40–50% while scaling from 3 to 150+ applications. • Fine-tuned Llama-3.2-3B (LoRA/QLoRA, Unsloth) as a routing/classification/redaction engine using 1,500 semantic prototypes from K-Means clustering over 50K production logs. • Established an LLMOps observability pipeline (MLflow, LangSmith) running asynchronously to keep latency minimal. • Containerized with Docker/Kubernetes on Azure; CI/CD via GitHub Actions.

Specialist Programmer - Strategic Technology Group at Infosys, Jan 2022 - May 2023 · 1 yr 5 mos, Trivandrum, Delivering Agentic AI systems for Apple’s Emerging Technologies & Services (ETS) org. AGENTIC TEST AUTOMATION & MCP INFRASTRUCTURE • Built custom MCP servers using generative AI for automatic Gauge specification generation from natural-language prompts, handling 500+ daily specification requests at peak load. • Built an LLM-powered test case/test data generation tool (LangChain), reducing manual specification authoring time ~70% across 20+ application modules; monitored via LangSmith. RAG PIPELINE & CODEBASE INTELLIGENCE • Architected a codebase-vectorization RAG pipeline (AST-aware chunking, HuggingFace all-MiniLM, FAISS, MongoDB Atlas Vector, Azure AI Search) with hybrid BM25 + semantic retrieval and RRF re-ranking. • Cut LLM token consumption ~60% across 15+ active repositories; added Responsible AI guardrails and metadata enrichment. • Built agentic AI chat interfaces (LangChain) for natural-language data retrieval from REST APIs and internal databases, adopted by 30+ engineers. DEPLOYMENT & INFRASTRUCTURE • Deployed production AI services (FastAPI, Flask) on Azure, containerized with Docker, orchestrated via Kubernetes, automated through Jenkins and GitHub Actions CI/CD — integrated with Azure ML and Azure OpenAI. • Applied prompt and context engineering to maintain sub-3-second response times at production load.

Digital Specialist Engineer - Strategic Technology Group at Infosys, Mar 2021 - Dec 2021 · 10 mos, Trivandrum, Infrastructure Demand Forecasting for Capacity Planning 100+ app teams on CVS's observability platform had no reliable method to anticipate infrastructure demand during high-traffic periods — risking budget waste or service degradation during Open Enrollment. • Designed and deployed an ML time series forecasting pipeline using SARIMA/SARIMAX to predict per-app infrastructure usage across 100+ applications, incorporating enrollment calendar regressors and seasonal demand patterns. • Delivered app-level forecasts replacing ad-hoc estimation with model-driven capacity planning, giving engineering teams ahead-of-time signals for resource allocation and de-allocation decisions ■ Adopted across full 100+ app portfolio · Directly informed infrastructure budgeting for a Fortune 500 healthcare client during their highest-stakes operational peri Failure Intelligence & ML Models Large-scale automated test executions generated hundreds of recurring failures daily with limited intelligent grouping or prioritization — increasing triage effort, slowing RCA workflows, and delaying engineering response times. • Developed predictive classification models using Scikit-learn to automatically categorize failures across 10+ failure types, achieving ~88% accuracy and reducing mean time to triage by ~35% — enabling optimized task assignment workflows • Implemented ML clustering algorithms for bulk failure analysis across 200+ daily test runs, identifying similar failure patterns across builds and cutting critical failure investigation time by ~40%. ■ ~88% classification accuracy across 10+ failure types · ~35% reduction in mean triage time · ~40% faster critical failure investigation across 200+ daily test runs

SDE Intern at Tech Mahindra, Jun 2019 - Aug 2019 · 3 mos, Noida Area, India, Developed UI and Validation tier elements using Angular and Java for e-TMR portal of the Government of India where users can file for patents and trademark registration.

Internship Trainee at i3indya Technologies, Jun 2018 - Jul 2018 · 2 mos, Delhi, Trained in Embedded systems (PIC microcontroller) and IoT platform.

Education

Bachelor of Technology - BTech, Electrical, Electronics and Communications Engineering, Maharaja Agrasen Institute Of Technology, Delhi

A.G.D.A.V.