Founding Engineer & AI Architect
ProPeers · July 2025 – Present
At a glance
- Architected the full AI ecosystem powering RoadmapAI, CodeLLM, AskAI, Global AI Search and the AI Code Editor building Agentic AI pipelines, RAG systems, MCP server architecture and LLM orchestration that now drives 80%+ of total platform traffic.
- Engineered RoadmapAI end-to-end with a self-learning RAG pipeline (text-embedding-ada-002, ChromaDB, semantic filtering, adaptive difficulty) and MCP-layered prompts, achieving sub-second inference and large-scale personalization.
- Delivered ~99% personalized roadmap accuracy using Agentic flows, structured prompt masks, multi-model routing, and RAG optimization directly improving RoadmapAI user ratings from the early 12% baseline.
- Built CodeLLM, an AI judge with multi-language detection, dual-layer JSON parsing, context-aware error classification (COMPILATION/RUNTIME/VALIDATION), semantic retrieval and deterministic verdict synthesis.
In the full detail
- Founding Engineer & AI Architect @ ProPeers - July 2025 – Present · Delhi, India · Remote
Founding Engineer & AI Architect
ProPeers · July 2025 – Present
Founding Engineer & AI Architect @ ProPeers - July 2025 – Present · Delhi, India · Remote
- Architected the full AI ecosystem powering RoadmapAI, CodeLLM, AskAI, Global AI Search and the AI Code Editor building Agentic AI pipelines, RAG systems, MCP server architecture and LLM orchestration that now drives 80%+ of total platform traffic.
- Engineered RoadmapAI end-to-end with a self-learning RAG pipeline (text-embedding-ada-002, ChromaDB, semantic filtering, adaptive difficulty) and MCP-layered prompts, achieving sub-second inference and large-scale personalization.
- Delivered ~99% personalized roadmap accuracy using Agentic flows, structured prompt masks, multi-model routing, and RAG optimization directly improving RoadmapAI user ratings from the early 12% baseline.
- Built CodeLLM, an AI judge with multi-language detection, dual-layer JSON parsing, context-aware error classification (COMPILATION/RUNTIME/VALIDATION), semantic retrieval and deterministic verdict synthesis.
- Developed AskAI, an agentic programming assistant using MCP-based prompt pipelines, resource-aware context analysis, dynamic O3Mini/O1 routing, token metering and automated formatting boosting engagement 3× and answer resolution speed 2×.
- Shipped the AI Code Editor with real-time AI review (<40ms), inline reasoning, multi-language execution and deep RoadmapAI/CodeLLM integration raising editor retention by 40%.
- Scaled Roadmap features to 120K+ organic users and improved MAU by 46% through rapid iteration, tight user-feedback loops and stable AI feature launches.
- Delivered Individual Roadmap Communities enabling peer-matching, shared progress tracking and roadmap-level micro-communities.
- Optimized CI/CD and deployment systems, cutting deployment time by 34%, automating multi-service rollouts, and enabling safer high-frequency releases.
- Reduced platform downtime by 90% (4 hrs to 45 mins/month) via infra hardening, progressive fallbacks, cache-first routing, real-time health checks and load-aware autoscaling.
- Implemented complete analytics & aggregation pipelines for 600K+ users with Redis caching, chunked batch aggregation, API acceleration and advanced rate-limit enforcement.
- Developed full search-validation engines (Roadmaps + RoadmapAI), ensuring context-safe retrieval, hallucination-resistance and consistent multi-node semantic validation.
- Performed Azure cost & infra optimization VM right-sizing, eliminated Bastion, stabilized Redis/Entra costs, contained Cognitive Service spikes and resolved large bandwidth egress surges.
Founding Engineer & AI Architect
ProPeers · July 2025 – Present
Founding Engineer & AI Architect @ ProPeers - July 2025 – Present · Delhi, India · Remote
- Architected the full AI ecosystem powering RoadmapAI, CodeLLM, AskAI, Global AI Search and the AI Code Editor building Agentic AI pipelines, RAG systems, MCP server architecture and LLM orchestration that now drives 80%+ of total platform traffic.
- Engineered RoadmapAI end-to-end with a self-learning RAG pipeline (text-embedding-ada-002, ChromaDB, semantic filtering, adaptive difficulty) and MCP-layered prompts, achieving sub-second inference and large-scale personalization.
- Delivered ~99% personalized roadmap accuracy using Agentic flows, structured prompt masks, multi-model routing, and RAG optimization directly improving RoadmapAI user ratings from the early 12% baseline.
- Built CodeLLM, an AI judge with multi-language detection, dual-layer JSON parsing, context-aware error classification (COMPILATION/RUNTIME/VALIDATION), semantic retrieval and deterministic verdict synthesis.
- Developed AskAI, an agentic programming assistant using MCP-based prompt pipelines, resource-aware context analysis, dynamic O3Mini/O1 routing, token metering and automated formatting boosting engagement 3× and answer resolution speed 2×.
- Shipped the AI Code Editor with real-time AI review (<40ms), inline reasoning, multi-language execution and deep RoadmapAI/CodeLLM integration raising editor retention by 40%.
- Scaled Roadmap features to 120K+ organic users and improved MAU by 46% through rapid iteration, tight user-feedback loops and stable AI feature launches.
- Delivered Individual Roadmap Communities enabling peer-matching, shared progress tracking and roadmap-level micro-communities.
- Optimized CI/CD and deployment systems, cutting deployment time by 34%, automating multi-service rollouts, and enabling safer high-frequency releases.
- Reduced platform downtime by 90% (4 hrs to 45 mins/month) via infra hardening, progressive fallbacks, cache-first routing, real-time health checks and load-aware autoscaling.
- Implemented complete analytics & aggregation pipelines for 600K+ users with Redis caching, chunked batch aggregation, API acceleration and advanced rate-limit enforcement.
- Developed full search-validation engines (Roadmaps + RoadmapAI), ensuring context-safe retrieval, hallucination-resistance and consistent multi-node semantic validation.
- Performed Azure cost & infra optimization VM right-sizing, eliminated Bastion, stabilized Redis/Entra costs, contained Cognitive Service spikes and resolved large bandwidth egress surges.