class YawarAbbas:
def __init__(self):
self.role = "Full Stack and AI Systems Developer"
self.education = "BS Software Engineering at UMT Lahore, 2023 to 2027"
self.company = "SE Intern at AgileTech Studio"
self.venture = "Cofounder of a Fiber ISP, grew from 20 to over 300 customers in 9 months"
self.building = "asmaan.com, a satellite first real estate marketplace"
self.focus = ["backend architecture", "agentic AI", "geospatial", "devops"]
def ships(self):
return "production systems that are deployed, monitored, and actually used"Before I wrote a line of Django, I was in the field, splicing fiber and knocking on doors in Choti Zareen. I convinced customers to pre pay before a single cable was laid, set up the server room, and configured the MikroTik routers. Growing an ISP from zero taught me more about execution and trust than any course could. Now I bring that same "make it actually work" mindset to code.
|
B2B SaaS, sole developer, live in production Real time parking facility intelligence platform. Ingests live IoT sensor data over MQTT, stores it in TimescaleDB hypertables, detects anomalies with Isolation Forest, and forecasts occupancy with Prophet, then pushes the results live to a dashboard over WebSocket. Built and deployed solo: 9 Docker services on Oracle Cloud, Nginx with SSL, RBAC, and 26 tests.
|
AI SaaS, solo build, live in production Job boards match on keywords, so you miss roles if you don't name the exact skill. Fyt reads a plain English description of who you are, infers your skills and level, scrapes live LinkedIn listings, and ranks every match from 0 to 100 with matching skills, missing skills, and an honest concern. Includes a CV Studio for ATS review and wizard based CV generation, plus Claude powered company research and cover letter tailoring.
|
|
LLM connected to real hardware, live demo An LLM agent that controls physical hardware through natural language. Say "the room's too hot, cool it down" and the agent reads live ESP32 sensor data, reasons over it, and flips a real relay that switches on a fan. A true feedback loop between a language model and the physical world, with tool calling, memory, guardrails, and voice control.
|
Self hosted TTS, backend and AI engineer Replaces the voice actor and studio loop for business phone systems. Clone a voice once, then generate unlimited IVR prompts from a dashboard. Self hosted Chatterbox TTS on GPU, a seven step approval workflow, and EBU R128 audio quality gates before anything goes live. Wrote a formal API contract together with the frontend engineer before writing any code.
|
|
Satellite first real estate, team lead Pakistan's first real estate marketplace where you browse properties on satellite imagery instead of stock photos. Leading a 3 person team: PostGIS geospatial queries, live Mapbox price pins, a listing verification state machine, and an XGBoost neighbourhood scoring model. A Jira backlog of more than 170 tasks, agile sprints, and a production stack.
|
Argus, the Agentic IoT Controller, and the IVR voice cloning system were built during my software engineering internship at AgileTech Studio. Fyt and asmaan.com are my own products. |
Languages
Backend and AI
Frontend
Data, Infra, and DevOps
Machine Learning Information Retrieval Software Quality Engineering Software Project Management Software Reengineering