Headline

AI Engineer & Researcher | IIT KGP EE Dual ’26 | Published at MICCAI, MedIA | Prev: OLA Krutrim, Harvard, IITGn | Deep Learning · LLM Systems · Medical Imaging

About

ML engineer and researcher graduated with a Dual Degree (B.Tech + M.Tech) in Electrical Engineering at IIT Kharagpur, specializing in Signal Processing and Machine Learning. I build ML systems end-to-end, from research to production. My work spans: • LLM orchestration at scale (Ola Krutrim, India's first agentic AI assistant) • Self-supervised and label-efficient vision models for medical imaging (MICCAI 2025, MedIA 2026) • Foundation models for hardware design (Harvard Edge Computing Lab, Prof. Vijay Janapa Reddi) • Multimodal learning for industrial fault diagnosis (M.Tech thesis, IIT KGP with GAIL) I've shipped a circuit-breaker LLM router handling multi-provider inference at production load, built adversarial and self-supervised frameworks for ultrasound landmark detection, and contributed to QuArch, the world's first computer architecture QA dataset. Open to ML engineering and ML research roles. Contact: anirvankrishna@gmail.com

Location

Patna

AI stack

antigravity

opencode

Work experience

Graduate Engineering Trainee at Ola Electric, Jun 2026 - Present · 3 mos, Bengaluru, Karnataka, India

Teaching Assistant at Indian Institute of Technology, Kharagpur, Jul 2025 - Apr 2026 · 10 mos, Mentored and evaluated UG and PG students across courses. • EE69205: Signal Processing System Design; Mentored a batch of 20 PG students, conducted continuous evaluation • EE22202: Algorithms, AI and ML Lab; Mentored 50+ students on concepts of statistical learning to deep learning

AI Engineering Intern at Krutrim, May 2025 - Jul 2025 · 3 mos, Bengaluru, Karnataka, India, Project: Kruti—India's first agentic AI assistant As an AI intern at Krutrim, Ola’s AI research and product division, I contribute to the backend architecture of Kruti—India’s first agentic AI assistant. My work focuses on building scalable, intelligent systems for large language model (LLM) orchestration, enabling multimodal, multilingual, and agent-driven AI services in production environments. • Designed and owned the full system architecture of a circuit-breaker-based LLM smart router—a core service powering Kruti, India's first agentic AI assistant. Wrote HLD and LLD documents that became the engineering spec. • Built a modular multi-provider LLM inference pipeline integrating Krutrim V2 and open-source models; added provider-specific streaming and Pydantic-validated structured outputs. • Implemented Redis-based distributed concurrency control across containers to eliminate race conditions in high-throughput routing. • Ran LLM-as-a-judge evaluations to benchmark quality, latency, and failure rates across providers; results directly informed routing policy. • Deployed and monitored production services via Docker, GitHub Actions, ArgoCD, Grafana, and New Relic.

Research Assistant at Indian Institute of Technology, Kharagpur, May 2024 - May 2026 · 2 yrs 1 mo, Master's Thesis: Multimodal Learning for Prognosis and Diagnosis of Bearing Faults Supervisor: Prof. Aurobinda Routray Researched and developed multimodal deep learning pipelines for automated prognosis and diagnosis of bearing faults—combining vibration signals, structured metadata and expert observation; robust to deployments in unseen conditions

Research Intern at Harvard University, May 2024 - Aug 2024 · 4 mos, Cambridge, Massachusetts, United States, Project: Generative AI for Hardware Design The work is focused on creating and labeling datasets in computer architecture, with an emphasis on data-centric and ML-based hardware design. The ultimate goal is to develop foundational models for hardware design. Supervisor: Prof. Vijay Janapa Reddi Work done: • Co-created QuArch, the world's first computer architecture Q&A dataset, spanning 50+ years of research literature—a resource for training and evaluating LLMs on hardware design tasks. • Led evaluation of GPT-4, GPT-4o, Llama 3-8B, and Mistral-7B on QuArch; achieved 94% F1, establishing the first benchmark baseline for architecture-domain LLMs. • Built a YAML-based evaluation framework for systematic parameter definition and pipeline integration, enabling reproducible LLM assessment at scale. • Contributed to the roadmap for a generative AI tool capable of autonomous hardware design tasks—including Verilog generation and microprocessor design.

Inventor- Invention Factory '23 at Indian Institute of Technology Gandhinagar, Jun 2023 - Jul 2023 · 2 mos, Gandhinagar, Gujarat, India, Invention Factory® India is a six-week summer program organised by the Maker Bhavan Foundation which originated in the USA in which IIT/NIT/BITS students from across India, working in teams of two, Prototype, Pitch and Patent inventions that each team has conceived of in the program’s first intensive week. • Innovated an Interconvertible Walking and Seating Aid for older adults who're prone to Neurocardiogenic Syncope • Conducted literature review regarding the heart rate and blood pressure profile during syncope to identify sudden bradycardia and orthostatic hypotension • Designed the actuation mechanism for automatic conversion of the device from walking to seating modes • Designed the algorithm for predicting the neurocardiogenic syncope, 2 minutes prior to its occurence • Recieved an honourable mention from Maker Bhavan Foundation for the device • Filed a Provisional Patent Application in India and the USA for the device (Provisional Patent Application no. 202311054748)

Education

B.Tech (Hons.): Electrical Engineering, M.Tech: Signal Processing and Machine Learning, Micro-Specialization: Artificial Intelligence and Applications, Indian Institute of Technology, Kharagpur

All India Secondary School Examination (AISSE), D.A.V. Public School BSEB Colony

Public links

github: https://github.com/Anirvan-Krishna