Hi, I'm
Mohan Krishna
G R
I build AI systems that |
Agentic AI Systems Researcher & Engineer — building multi-agent coordination protocols, causal GraphRAG systems, and edge AI pipelines deployed in aerospace, fintech, and safety-critical environments.
DAR3 Protocol — Multi-Agent Coordination State Machine
Research Focus
Agentic AI & Multi-Agent Coordination
DAR3 Protocol · CertusHire · 2Cents Capital
“How do AI agents coordinate under uncertainty in real-time systems?”
- ▹85% agreement with human evaluators (CertusHire)
- ▹<200ms response latency at 100+ concurrent sessions
- ▹DAR3: novel state-machine coordination protocol for LLM agent orchestration
Graph-Augmented & Causal Reasoning
GraphRAG · Causal XAI · Airbus
“Can causal graphs make LLM reasoning interpretable and trustworthy?”
- ▹25% higher retrieval accuracy vs. dense-only baselines (Airbus)
- ▹Causal counterfactual perturbation studies for XAI validation
- ▹Top 5 Paper Award, Airbus TechXceed 2025 (350+ engineers)
Efficient ML & Edge Deployment
TensorRT · Quantization · PEFT · Edge AI
“What are the limits of deploying large models under hard latency and power constraints?”
- ▹90% inference speedup via TensorRT FP16 on Jetson Nano (30 FPS at 720p)
- ▹2× throughput improvement with quantization + PEFT at Airbus (negligible accuracy loss)
- ▹86.7M-parameter model optimized for <10W edge deployment
Open Research Questions
Questions I'm actively thinking about — the kind that don't have clean answers yet:
- Q1.
“How do we formally verify correctness properties of multi-agent coordination protocols like DAR3?”
- Q2.
“Can causal graph conditioning replace chain-of-thought prompting for interpretable LLM reasoning?”
- Q3.
“What are the fundamental latency limits of multi-agent inference at production scale?”
- Q4.
“How do sparse retrieval signals generalize across domain-specific corpora (aerospace vs. fintech vs. medical)?”
Experience & Journey
Sep 2022 – Mar 2027
Master of Technology, Computer Science & Engineering (5-Year Integrated)
Sri Ramakrishna Engineering College
Coimbatore, India
Jan 2026 – Feb 2026
AI Engineer Intern
2Cents Capital · Dubai, UAE (Remote)
Multi-Agent Systems & Asset Management Research
- ▹Engineering Agentic AI systems (Valura.ai) for investment research and asset-management decisioning
- ▹Researching multi-agent coordination under financial domain constraints
Jun 2025 – Dec 2025
Generative AI Intern
Airbus · Bengaluru, India
GraphRAG, Explainable AI & Transformer Optimization
- ▹Proposed GraphRAG variant achieving 25% higher retrieval accuracy over dense-only baselines for long-context LLM reasoning on engineering corpora
- ▹Designed causal graph–conditioned RAG framework for interpretable root-cause analysis, validated via counterfactual perturbation studies
- ▹Transformer inference optimization via quantization + PEFT: 2× throughput with negligible accuracy loss
- ▹Top 5 Paper Award at Airbus TechXceed 2025 (AgenticEye) among 350+ engineers
- ▹Selected speaker at Airbus Global AI Week 2025
Jan 2025 – Mar 2025
Trainee Software Engineer
SuperDNA 3D Lab · Hyderabad (Remote)
Computer Vision & 3D Reconstruction
- ▹CV pipelines for automated QA of AI-generated e-commerce imagery (15-rule rubric)
- ▹Dimension estimation in monocular images using depth transformers
- ▹Experimental 3D hair reconstruction with UniHair
May 2024 – Jul 2024
AI/ML Intern
Infosys Springboard · Remote
NLP & Transformer Fine-tuning
- ▹TextSumm: empirical comparison of extractive vs. abstractive summarization on a 564K+ document corpus
- ▹3× ROUGE-2 improvement using fine-tuned transformers combined with TF-IDF/KMeans heuristics
- ▹Deployed reproducible experiments on Azure ACI with CI/CD
$ about --me
Research Projects
2025 · Featured Project
CertusHire
Autonomous Multi-Agent Interview Platform
“How do AI agents coordinate under uncertainty in real-time adaptive evaluation systems?”
- ▹Designed DAR3 — a state-machine–based multi-agent coordination protocol enabling dynamic Interviewer ↔ Evaluator role assignment based on real-time candidate performance
- ▹Implemented hybrid dense–sparse RAG pipeline (embeddings + TF-IDF) achieving 85% agreement with human interviewer judgments
- ▹Scaled to 100+ concurrent sessions with fault-tolerant FastAPI + Celery/RabbitMQ backend at <200ms latency
85%
Human Agreement
<200ms
Response Latency
100+
Concurrent Sessions
90%
Screening Effort Reduced
2024 · Featured Project
PyroGuardian
Edge-AI UAV Fire Detection System
“How do we deploy an 86.7M-parameter vision model on a <10W edge device at 30 FPS?”
- ▹Applied TensorRT FP16 + CUDA acceleration achieving 90% inference speedup vs. PyTorch FP32, sustaining 30 FPS at 720p
- ▹Curated 51GB multi-scenario dataset — 10K+ annotated frames across 8 fire conditions with targeted augmentations
- ▹AWS SNS alerts with RBAC roles: <1s latency for 500+ users, dynamic severity-score prioritization
90%
Inference Speedup
30
FPS at 720p
51 GB
Dataset Size
<1s
Alert Latency
Other Notable Projects
“Can hybrid extractive–abstractive pipelines outperform pure transformer baselines at scale?”
“Can quantum annealing solve combinatorial agricultural scheduling faster than classical solvers?”
“Can a hybrid ML model detect stress states accurately enough for clinical-grade mobile deployment?”