Headline

AI Engineer | LLM Agents, RAG & Evaluation | Ex-Microsoft (2x Microsoft Impact Award) | IIT Kharagpur

About

Software engineer with 1+ year at Microsoft building agentic compliance workflows and browser test infrastructure, plus end-to-end AI systems I built on my own: multi-tenant RAG, lead triage, an operations agent and a voice receptionist. I focus on measurable quality. My RAG platform reaches 0.95 faithfulness with zero breaches across 83 red-team attacks, and my operations agent had 0 forbidden actions across 640 runs. What I work with: Python, FastAPI, LangGraph, MCP, PostgreSQL, Docker, Langfuse. Dual Degree, IIT Kharagpur, with a micro-specialisation in Artificial Intelligence & Applications. Open to AI Engineer roles in Bengaluru or remote. Email: anushkakulkarni527@gmail.com | GitHub: github.com/Anushka-200

Location

Bengaluru

AI stack

claude code

hugging face

langchain

langfuse

langgraph

langsmith

Work experience

Software Engineer at Microsoft, Jul 2025 - Sep 2026 · 1 yr 3 mos, LLM Agentic Compliance Plugin (Edge) • Goal: replace manual multi-reviewer compliance sign-off on Edge pull requests with an LLM agent that reviews and fixes code against 5 tenets, and never acts without approval. • Built stage-aware orchestration for the plugin across five review tenets, integrating Azure DevOps APIs and human-approval gates. • Orchestrated drift detection across three compliance axes, with resolve/override handling and automated PR audit tags. • Achieved 100% pass across 13 automated evaluations (47 decision points) via a 4-tier regression suite; adversarial tests exposed and helped close a confirmation-gate bypass. Virtual Wallet and Digital Credentials Test Infrastructure • Goal: make the W3C Digital Credentials API testable in CI, which previously needed a real wallet app and a human tapping through each test. • Built a simulated wallet enabling CI/CD automation; owned 2-layer test infrastructure across ChromeDriver, Chrome DevTools Protocol and WebDriver BiDi. • Extended CDP with a Digital Credentials domain and built WebDriver BiDi support (protocol definitions, routing, schema validation); contributed upstream to Chromium; added Perfetto tracing. Operational excellence • On-call DRI: 7 production incidents mitigated, 30+ bugs triaged, 150+ build failures resolved across domains.

Data Science Intern at Microsoft, May 2024 - Jul 2024 · 3 mos, Hyderabad, Telangana, India, • Developed predictive URL feature in Microsoft Edge using BERT model for enhanced user engagement • Engineered global evaluation metric, reducing reported accuracy from 37% to 15% • Implemented incremental training of BERT on new datasets for 1-2% accuracy improvement • Updated model's understanding of new domains by modifying vocabulary using tokenizer and transformer libraries

Contingent Member | Inter-IIT Cultural Meet 6.0 at Technology Students'​ Gymkhana, IIT Kharagpur, Nov 2023 - Dec 2023 · 2 mos

General Secretary Students' Welfare at Sister Nivedita Hall of Residence ,IIT Kharagpur, Oct 2022 - Aug 2023 · 11 mos, • Implemented a wing representative system to support 1148 students at Sister Nivedita Hall of Residence, IIT Kharagpur. • Organized celebratory events and engaging group activities to foster a positive atmosphere for students. • Conducted informative fundae sessions by experienced seniors to create a conducive learning environment. • Established a dedicated students’ welfare room for recreational activities, enabling students to relax and bond with their peers

Student Member at Technology Dance Society, IIT Kharagpur, Dec 2021 - Jan 2024 · 2 yrs 2 mos

Education

Dual Degree, Mechanical Engineering, Indian Institute of Technology, Kharagpur

10th, Nashik Cambridge CBSE school

Public links

github: https://github.com/Anushka-200