Loading projects…
// projects
From an AI tutor used by 600,000+ learners to agentic platforms, cognitive-security research and sovereign AI: what each system does, the result it delivered, and where to see it run.
17 projects
Techmatic Systems
Visual AI workflow and agent platform: build on a canvas, edit any node's Python, and serve the same flow as REST, MCP or an OpenAI-compatible API.
100% pass rate across 400 automated and manual test cases
Proves an acknowledged write really happened: a bounded-poll read-back that checks existence, state, exact values and parent links within a per-backend budget.
Catches 4/4 silent corruptions that == and is-not-None miss; 0 overruns in 2,000 runs
A proprietary 3.1B-parameter transformer pretrained from scratch for drug law enforcement.
3.1B-parameter model pretrained from scratch with a custom tokenizer
Detects human-targeted cyber attacks as measurable distortions in trust networks, not as malicious text. Built solo in a 41-hour sprint.
ROC-AUC 0.9555, ahead of every text-based baseline (best: 0.9283)
0 0updated Mar 2026
Sovereign OSINT identity intelligence for law enforcement and threat-intelligence teams. DPDP Act 2023 compliant.
Correlates digital presence across 2,003+ platforms in under 60 seconds
1M-turn multi-agent negotiation dataset for India's kirana economy: informal Udhar credit, supply-chain signals and multilingual bargaining dialogue.
1,000,000 negotiation turns across 125,000 trajectories in 12 states
0 0updated Jun 2026
Real-time AQI monitoring and forecasting for 15 Hyderabad stations, with five ML models, health advice and PDF reports; the system behind the AQI paper.
RMSE 3.7 across 15 monitoring stations (multi-modal deep learning)
AI code-generation and tutoring platform: plain-language prompts become optimised, explained code in 20+ languages, with live preview.
Scaled to 600,000+ enrolled learners
AI career and education platform (formerly QwiXCV): resume builder, ATS scanner, interview coach, career simulators and blockchain-verified credentials.
Outcome (measurement pending)22 tools across CV, career, learning and Web3 suites
0 0updated Dec 2025
Execution-aware, multi-agent Text-to-SQL: schema GraphRAG, a local LLM ensemble, safe execution and a semantic critic, running fully offline.
34% zero-shot execution accuracy on unseen Spider schemas, fully local
5 0updated Mar 2026
A cognitive mobility engine that models human intent, identity and trust as evolving cognitive states, to help people work across borders.
Outcome (measurement pending)Six-layer cognitive stack (NCA-1) on Gemini 2.5 Flash
0 0updated Dec 2025
A decision cognition runtime for Tableau Cloud: live insight confidence, decision intent and decision risk, built with native calculations and no AI.
Outcome (measurement pending)4 coordinated dashboards for confidence, intent, risk and memory
0 0updated Dec 2025
A self-optimizing compiler that encodes IR with a Transformer and learns optimization rules with RL.
0.28 execution-cost reduction vs. a 0.00 LLVM baseline
A cyber-manipulation benchmark covering phishing, OTP fraud and social engineering, with a Cognitive Manipulation Score.
100,000 samples scored with a Cognitive Manipulation Score (Apache 2.0)
AI pronunciation analyzer for American English: records your voice and gives real-time feedback on vowels, consonants, stress, rhythm and intonation.
Outcome (measurement pending)Real-time feedback across vowels, consonants, stress, rhythm and intonation
0 0updated Jul 2025
Restaurant web app for Dine 24, Hyderabad: table booking with pre-orders, an AI dining assistant, UPI payment, emailed PDF bills and an admin portal.
Outcome (measurement pending)Live: reservations, pre-orders, AI assistant and an admin portal for menus, specials and coupons
Calculates your ecological footprint from food, housing and transport habits, and shows the CO₂ equivalent as cigarettes burnt to make it relatable.
Outcome (measurement pending)CO₂ equivalents translated into cigarettes burnt
0 0updated Jul 2025
From research prototype to production system: tell me what you need to ship.