B.Tech Computer Science (AI & ML) · MITS 2023–27
LAKSHMI SAGAR
SESHADRI
I operate at the intersection of deep research and production engineering — building systems that don't just prototype ideas, but scale them.
Engineering lightweight neural networks, geospatial intelligence pipelines, and ensemble defense architectures.
01 / SYSTEMS
Deployed Systems
Production & Research Scale
Lightweight Feature Attention Network (LFAN)
Built a lightweight CNN for ×4 single-image super-resolution achieving 29.09 dB PSNR / 0.806 SSIM on DIV2K. Improved all 100/100 validation images over bicubic baseline through iterative architecture tuning.
LLM Inference Engine
From-scratch continuous batching server re-implementing vLLM's core. PagedAttention, Triton attention kernel, autonomous ReAct scheduling agent.
UCIS
Urban Cooling Intelligence System. Ingests Sentinel-2 & Landsat-8, runs geospatial hotspot analysis via PostGIS on free-tier cloud infra.
CropGuard AI
Full-stack AI system for crop disease detection using image analysis, supporting multiple crops and diseases.
EthosForge
Ensemble classifier for financial fraud detection with strict focus on ethical AI and model interpretability.
SoilTwin
Full-stack digital twin providing real-time analytics and ML-driven insights on soil characteristics.
AssumptionX
AI-powered web app identifying and challenging hidden assumptions in business plans. Deployed at Vercel.
Xenovex Technologies
Web Dev Intern. Delivered full-stack Mood Journal with JWT auth, C# .NET API, and PostgreSQL.
02 / TIMELINE
Experience & Milestones
Industry
Web Development Intern
Xenovex Technologies
Delivered full-stack Mood Journal application with JWT authentication, C# .NET backend API, and PostgreSQL database.
Inceptrix 2.0 @ JAIN University
KrishiMind · Soil Intelligence System
Built KrishiMind — an AI-driven agricultural intelligence system with soil digital twin capabilities at national-level hackathon.
Academic
B.Tech CSE (AI & ML)
Madanapalle Institute of Technology & Science
GATE DA — AIR 9277
Graduate Aptitude Test in Engineering
Qualified GATE DA. Targeting sub-1000 AIR in GATE 2027 — a serious, non-negotiable goal.
Active Research Pipeline
3 manuscripts · peer review
Geospatial CV, LLM systems, and scientometric studies. Springer LNNS oral presentation accepted for ICTIS 2026, Bangkok.
03 / STACK
LANGUAGES
Python
TypeScript
Java
C
C#
React.js
Next.js
Angular
FastAPI
PostgreSQL
GCP
VercelPythonTypeScriptJavaCC#React.jsNext.jsAngularFastAPIPostgreSQLGCPVercel
PyTorch
TensorFlow
OpenCV
Scikit-learn
Docker
Supabase
MongoDBPyTorchTensorFlowOpenCVScikit-learnDockerSupabaseMongoDB04 / RESEARCH
Academic Literature
A Review on Large Language Models in Engineering
Springer Lecture Notes in Networks and Systems (LNNS). Paper ID 1235. Oral presentation, Bangkok.
Characterizing Small-Scale Deforestation Trends in the Northeastern Brazilian Amazon
Landsat NDVI Time Series · Remote Sensing
Mapping the Evolution: A Scientometric Review of GIS Applications in Flood Hazard Mapping
1985–2025 · Bibliometric Analysis
Geospatial AI
Satellite imagery analysis, NDVI temporal studies, urban heat island detection, flood hazard mapping with PostGIS.
LLM Systems
LLM engineering review, RAG failure characterization under temporal corpus drift, agentic architectures.
Digital Twins
Adaptive digital twin frameworks for smart infrastructure, hybrid quantum-classical DT architectures.
05 / RECOGNITION
Awards & Recognitions
Springer LNNS Publication
Accepted · Oral Presentation
Paper ID 1235 accepted for oral presentation at ICTIS 2026, Bangkok. Published in Springer Lecture Notes in Networks & Systems.
GATE DA Qualified
AIR 9277 · 2026
Qualified the Graduate Aptitude Test in Engineering — Data Analytics stream. Targeting sub-1000 AIR in GATE 2027.
Inceptrix 2.0 Hackathon
JAIN University · National Level
Built KrishiMind — a soil intelligence & digital twin system — at JAIN University's national-level hackathon.
29.09 dB PSNR — LFAN
Super-Resolution · DIV2K
Achieved 29.09 dB PSNR / 0.806 SSIM on DIV2K with only 731K parameters. Improved 100/100 test images over bicubic.
DisasterMesh
Hackathon · Disaster Response
Developed DisasterMesh — a mesh-network-based disaster response coordination system for offline connectivity scenarios.
CGPA 8.67 / 10.0
MITS · Year 3 of 4
Maintaining 8.67 CGPA while running multiple production projects, active research pipeline, and enterprise sales role.
06 / CONTACT
Initiate
Contact
Open for discussions regarding ML infrastructure, system architecture roles, and academic collaboration.