I'm a backend engineer crossing over into AI infrastructure. I've spent the last year building automation pipelines — moving data between APIs, orchestrating workflows, shipping features that serve real users. Now I'm going deeper: understanding the math behind ML, building systems that wrap and serve models, and learning how production AI actually works.
I don't just want to use AI tools. I want to understand them, build with them, and eventually design the infrastructure that makes them reliable.
Bluebeaks Solutions — AI Engineer Intern (May 2026 - Aug 2026)
Built the AI layer of a resume tailoring platform. My work sat at the intersection of backend engineering and ML operations:
- Designed an n8n pipeline that ingests job descriptions, scores resumes against them using Gemini LLM prompts, and outputs tailored PDFs — processing 200+ resumes/week at 95%+ formatting accuracy
- Built a multi-persona drip email system with five user segments and cron scheduling, cutting manual outreach by ~70%
- Scaled lead generation to 400+ companies across 5 countries, enriching 10,000+ contacts via Tomba and Cleanlist APIs — 3x prospecting efficiency
- Wrote a Python transformation layer handling 50+ API payloads/day at 99.9% data integrity, supporting multi-step workflow orchestration
- Integrated Gemini with guardrail prompts to generate structured outputs for 100+ job descriptions/month, reducing human review by 40%
- Built automated job-application workflows using Zerowork, saving ~6 hours/week of manual work
Key realization from this role: shipping AI features isn't about the model — it's about the pipeline around it. Data integrity, prompt guardrails, fallback logic, and observability matter more than the LLM itself.
Journald-Sniffer (2025 - Present)
Linux auth event ingestion and threat detection pipeline. Reads systemd journal (facility 10 — authpriv), parses sudo/su/sshd sessions, classifies outcomes, escalates ambiguous ones to Groq LLM, and alerts on suspicious patterns. Exposed via FastAPI and containerized with Docker.
- Pipeline:
ingestor.py→parser.py→llm.py→watchdogv2.py - API: FastAPI with 6 endpoints —
/health,/ingest,/parse,/alerts,/sessions,/raw - Classification: success / failure / suspicious / unknown via keyword matching + Groq LLM fallback for ambiguous sessions
- Threat detection: brute-force (5+ failures from same IP), port scan (12+ neutral events, 0 successes), success-after-failure
- Storage: PostgreSQL with three tables —
raw_logs,auth_logs,ingest_state - LLM: Groq
llama-3.1-8b-instant— only called for ambiguous sessions to minimize API costs - Deployment: Dockerfile + docker-compose, uvicorn ASGI server
RepoRecon (2026)
Automated repo structure analysis and code intelligence. Currently rebuilding it with:
- AST parsing for Python codebases
- Embedding-based code search (moving from keyword to semantic)
- This is my playground for learning how vector search actually works
skill_sYnc (2026)
Flutter app connecting users based on shared skills. Firebase backend, production-deployed. My reminder that shipping matters more than perfection.
MediScribe (2026)
Medical documentation assistant built with Dart + Flutter. Explored how LLMs can be constrained to produce reliable clinical text — mostly learned what doesn't work.
- 400+ LeetCode problems solved (max rating 1616)
- 200+ resumes/week automated at Bluebeaks
- 10,000+ contacts enriched across 5 countries
- 6 REST endpoints in Fac-10Sniffer
- SIH 2025 Grand Finalist — ISRO problem statement
- GDG On Campus Core Technical Team
Languages: Python, SQL
Backend / APIs: FastAPI, Django REST Framework, REST APIs
Automation / AI: n8n, Google Gemini API, Groq API, RenderCV
Databases: PostgreSQL, MongoDB
Networking: TCP/IP, DNS, DHCP, HTTP/HTTPS, SSH, IPv4/IPv6, Routing, Subnetting
Core CS: Data Structures & Algorithms, Operating Systems, DBMS,
Computer Networks, OOP, Software Development
Tools: Git, GitHub, Docker
Currently seeking: 2027 SWE internships in AI infrastructure or backend engineering. Open to relocating. Let's talk: ootsodhar@gmail.com