About
B.Tech CSE (AI) at Amrita. Ship AI systems and low-latency infra. Graduating 2027.
// Builder · Engineer · Designer
Keerthivasan S V
Keerthivasan. Third-year B.Tech AI student. I ship agentic AI systems and high-performance infra. Shipped a voice-agent platform live (SpinSci Grand Prize, ₹1L), won 1st at Hack Web3Conf 2025 ($1K, FHE governance), and hold an approved patent in IoT computer vision.
Currently Founding Builders Cohort at Smartly Ventures, App Dev plus PM Intern at Navicon Infraprojects (10+ months), and just wrapped Cohort 6 at Protocol Labs Dev Guild (libp2p interop).
Where I trained.
- Coursework: DSA, Probability & Statistics, Linear Algebra, ML Fundamentals, System Design, Computer Networks, Big Data Analytics, CV & Image Processing.
- Specialised tracks: applied AI for production, agent reliability, computer-vision deployments.
- Focused on Mathematics, Physics, Biology, Chemistry, and Languages.
- Strong investment in தமிழ் literature alongside English. Still write in both.
Where I shipped.
- Early team at an applied AI venture studio. Agent infra and real-world deployments for emerging markets.
- Core build phase: WhatsApp business automation for MSMEs (salons, clinics). Lead generation, booking, retention, promotion.
- Dogfooding internal GTM systems (CSR outreach, partnerships, college outreach) to harden the agent infra before external deployment.
- Founding cohort, building toward a $6M SAFE round with plans to extend to $10.5M seed.
- 12 weeks on
libp2p/unified-testing, the interop + scaling framework for libp2p implementations. - Resolved 8 critical bugs in the js-libp2p v3.x CI pipeline. Expanded Docker test orchestration across 49+ protocol combinations.
- Cross-implementation interop validation across Python, Rust, JS, Go nodes.
- PM Intern (Nov 2025 to Present): working with the Director on strategy and delivery for a 9-module ERP platform. 6-engineer team, sprint coordination, KPI tracking (adoption, uptime, feature usage, CSAT).
- P95 latency under 1.4 s and 99.5%+ uptime for 10+ active users in production across Android and iOS.
- App Dev Intern (Jun to Nov 2025): built the Flutter ERP app with role-based auth and Supabase real-time sync. AI Master Data Agent (FastAPI plus DistilBERT plus spaCy), 35 intents, 15+ entities, 90%+ accuracy.
- Multilingual OCR plus invoice parsing (85%+ F1), Prometheus and Grafana observability, JWT-based RBAC across 100+ endpoints.
What I reach for by default.
- C++
- Python
- TypeScript
- Solidity
- C
- Low-latency design
- Lock-free structures
- FIX protocol
- Statistical arbitrage
- Order books
- Backtesting pipelines
- PyTorch
- TensorFlow
- DistilBERT
- spaCy
- OpenCV
- ONNX
- pgvector
- YOLOv11
- MediaPipe
- Whisper
- Gemini Flash
- Ethereum
- Optimism L2
- Shardeum
- Inco FHE
- Hardhat
- Ethers.js
- Smart Contracts
- Docker
- Kubernetes
- PostgreSQL
- Redis
- Supabase
- Prometheus
- Grafana
- FastAPI
- Next.js
- Flutter
- AWS Bedrock
- E2B Sandboxes
- MCP
- Deepgram
- ElevenLabs
- AWS Bedrock
- Figma
Work stamped by the system.
IoT-Based Edge AI Pipe Inventory System
Patent No. 202541075966. YOLOv11 on Raspberry Pi 4. 98.7% accuracy on single pipes, 82%+ on 4-layer nests. Sub-200 ms inference via model pruning plus TensorRT.
ApprovedEdge-Deployable Pipe Counting via Hybrid CV
Companion IEEE paper documenting the methodology behind the patent. Hybrid deep learning plus classical CV for industrial pipe inventory, nested-pipe segmentation, model quantisation.
Under reviewSIOF · Semantic Integrity & Orchestration Framework
Python 3.14 free-threading toolkit on PyPI. Parallelised AST indexer (PEP 703), De-Slopper engine for AI-code anti-patterns, secure MCP server with RBAC, Green Guard energy tracking.
242 tests · liveAI/ML in Pyrolysis Systems
Review work supervised by Dr. Akhil Mohan (CEN-ASAI Amrita / KTH). Physics-informed neural networks, predictive modelling, autonomous pyrolysis system design.
In progress