Anshul
Raman
Security-minded builder. I write software with an adversarial lens — studying how systems break so I can build ones that don't. From full stack platforms to offensive labs, I build things that are meant to last, driven by a deep interest in AI and a curiosity for how systems can be tested, broken, and rebuilt stronger.
Who I am
and why it matters.
A kid who broke
everything.
I grew up taking apart phones, routers, and toys — not because I hated them, but because I wanted to see how they failed. That same curiosity now drives my work across full stack development, AI systems, cybersecurity, networking, and systems programming.
Build fast,
break it safely.
Short feedback loops: I prototype, break, and harden systems. At UC, I work on full stack and AI projects alongside offensive security labs, hardware builds, and Python tools to understand where real systems crack, from Wi-Fi to phishing to local networks.
What keeps me
moving.
Clubs: Tennis Club · Bearcat Tech Innovators · IEEE · AAEO · UC Robotics · AI Society
Sports: Tennis · Badminton · Cricket · Soccer · Volleyball
Currently
at Cincinnati.
Freshman at the University of Cincinnati studying Computer Science with a focus on full stack development and AI systems, building production-ready web applications and user-focused software, with growing experience in cybersecurity through hands-on labs and systems exploration. Dean's List Fall 2025. Open to summer roles and co-op opportunities starting Fall 2026.
Where I've
shown up.
Alipur, Delhi, India
- Led and managed school-wide science events, demonstrating leadership, professionalism, and team coordination.
- Collaborated with staff and student teams to design interactive STEM activities, fostering a culture of curiosity and innovation.
Karnal, Haryana, India
- Supported children with special needs using patience, empathy, and adaptive communication techniques.
- Assisted children with hearing and speech impairments through tailored learning approaches.
- Coordinated an inclusion awareness event at Pratap Public School to promote social integration.
- Collaborated with staff and volunteers to ensure effective event execution and engagement.
Things I've
actually built.
3rd Place · Medpace Clinical Trial Challenge @ RevolutionUC 2026 · MLH Best Use of Presage API · Built in 24 hours.
- Built a full-stack AI clinical intelligence platform in 24 hrs — voice check-ins, real-time physiological sensing (heart rate, breathing rate, stress) via the Presage API camera SDK, medication scanning with AI safety and dosage analysis, and voice-phrase authentication.
- Engineered three role-based dashboards: patients interact entirely through voice; doctors get Gemini-powered risk scoring that auto-flags patients as stable, moderate, high risk, or critical; lead doctors get a global command center with doctor performance analytics, cohort insights, and trial-wide adherence and dropout metrics.
- Built a simulation lab for generating synthetic patients and modeling clinical workflows, plus an optimization engine predicting trial completion likelihood and identifying at-risk individuals before dropout.
- Integrated real-time emergency alerting, AI symptom matching to patient history, built-in doctor-patient messaging, and voice-activated login with exact-phrase verification — any deviation fails authentication.
Full-stack GPA dashboard with AI assistants, grade simulator, study planner, and Canvas LMS integration. Live at gradintel.anshulraman.com
- Cloud-synced GPA dashboard: cumulative GPA, semester GPA, credits, at-risk warnings across every device via Supabase.
- Grade Simulator with sliders for what-if scenarios + safe-zone calculator for every remaining exam.
- Multi-model AI Assistant — Gemini, Grok, Claude, OpenAI — plus syllabus PDF parser that auto-populates exam structure.
- Study Planner: Pomodoro timer, deadline tracker, mood log, AI-generated risk-ranked focus plans.
- Canvas LMS integration to auto-pull live grades from the university portal.
- Degree Tracker with AI-powered degree-plan PDF import and target GPA calculator.
- Gamification: XP, badges, GPA streaks, level titles, shareable cards, configurable GPA scales (4.0 / 4.3 / 5.0).
Full-stack cold email automation that finds recruiters across 4 job boards, generates personalised emails, and sends them from your own Gmail. Built for job seekers who want to skip the queue.
- Cloud-synced campaign dashboard: emails sent, companies reached, delivery rate, and campaign history across every device via Supabase.
- Multi-source job search across LinkedIn, Indeed, Glassdoor & Handshake — filter by source and select listings in one click.
- Recruiter email finder that generates 6 pattern variations per contact (
firstname@,f.lastname@,fi.last@, etc.) with confidence scoring. - Variable-driven email template editor with live preview — auto-swaps recruiter name, company, job title, and your skills per send.
- Gmail & Outlook OAuth integration: emails sent from your own account with resume attached, never from a shared server.
- 4-step campaign wizard: profile builder, job selector, email composer, and live send queue with per-job status tracking.
- Setup integrations for Hunter.io (email accuracy) and RapidAPI JSearch (live job data) with embedded step-by-step instructions.
Fully working AI phone agent that answers calls and holds real conversations, running completely free on a local machine.
- Voice pipeline integrating Twilio call handling, local Whisper transcription, LLaMA 3.2 via Ollama, and gTTS text-to-speech into a single real-time conversation loop.
- Dual server architecture with Node.js handling Twilio webhooks and Python handling AI processing, communicating internally to keep the call alive during inference.
- Built a polling system to solve live call timing — keeps the call open every 5 seconds while AI processes, then plays reply the moment it's ready.
- Fully local and free — no OpenAI, no paid APIs. LLaMA 3.2 runs on device via Ollama, Whisper transcribes locally, total cost is $1/month for the Twilio number.
- Debugged and resolved webhook routing, call state management, Python–Node intercommunication, and package dependency conflicts across the full stack.
Fully autonomous robot that navigates a line track, detects and grabs a bin by weight classification, and delivers it to the correct drop zone. Built from scratch with zero prior hardware experience.
- Single LabVIEW While Loop with a GRabbed boolean state machine controlling every behavior in real time — line following, bin detection, claw actuation, carry, and drop-off — no sequential structures, no human input.
- Dual color sensors in Detect mode straddle the 1-inch black line edges for real-time directional correction. Ultrasonic proximity sensor triggers motorized gear-driven claw at under 5cm distance.
- 4 physical prototype iterations (gear train → tricycle → forklift redesign → final claw build) and 5 complete software architecture rewrites before landing on the final design.
- 45 documented test runs across 3 experimental factors — surface color, lighting conditions, and tape contrast — with full descriptive statistics to validate sensor reliability before demo day.
- Applied the full engineering design process: stakeholder analysis, morphological charts across 6 design functions, weighted decision matrices, and descriptive statistics — not just built it, actually engineered it.
Full console-based banking with persistent storage, PIN security & admin panel.
- Unique account IDs and PINs with duplicate-collision detection.
- Persistent storage via
save_accounts()— rewrites accounts.txt after every operation. - 3-attempt PIN lockout with auto-lock on failure, reset on success.
- Admin panel: view all accounts, unlock cards, robust input validation.
File-backed bus reservation with seat tracking, cancellation, and unique booking IDs.
- Unique booking IDs validated against existing records to avoid collisions.
- Per-bus seat availability synced on every booking and cancellation.
- File storage persists bookings, rebuilding seat counts at startup.
- Handles invalid input gracefully; rebuilds bookings file after cancellations.
Assistive hardware for Alzheimer's and dementia patients — reduces missed doses and improves caregiver visibility.
- 4-person team: smart dispenser verifies patient, delivers correct dose, escalates emergencies.
- Camera-based facial recognition before dispensing — zero wrong-patient risk.
- Automated alerts to caregivers + emergency call trigger in life-threatening scenarios.
- Addresses medication adherence, caregiver burden, and patient safety.
Controlled Remote Access Trojan lab — full-device compromise, 12+ remote functions, ethical documentation.
- Cross-platform RAT: real-time surveillance, GPS tracking, SMS exfiltration in isolated lab.
- Post-exploitation research on Android & Windows — documented system permission weaknesses.
- Compiled ethical-use guidelines and mitigation steps as final deliverable.
ESP8266 wireless auditing — evil twin APs, deauth testing, 50+ beacon floods.
- Portable tool evaluating 802.11 vulnerabilities and deauthentication resilience.
- “Evil Twin” APs audited 20+ device connection attempts in a controlled environment.
- 50+ simultaneous beacon signals to study wireless congestion thresholds.
End-to-end phishing campaigns in a secure lab — credential harvesting, MFA mitigation.
- Rebuilt phishing workflows to study password capture and social engineering tactics.
- Fake login portals and credential harvesting flows recreated in secure local lab.
- Mitigation: MFA enforcement, user awareness training, email filtering policies.
What I actually
reach for.
Not a buzzword list. These are the tools I use when things get real.
Where I'm
learning.
- GPA 3.68 · Dean's List Fall 2025
- Focus: Cybersecurity & Software Engineering
- Data Structures · Algorithms · Systems · Security Labs
something real.
I'm always open to interesting problems, co-op opportunities, or a good conversation about security. Best way to reach me is email.