Lily’s Operating System
“everything is a system, and this is mine.”
IMG_RESRCH_NODE_09 // ACTV
AI in bioinformatics, astronomy, sustainability
Why do I build? Statement of purpose and active research domains exploring gene alignments, astronomy data grids, and resource sustainability.
CRIS (Centre for Railways & Information Systems)
National scale AI chatbot for multiple Civil Engineer Projects in Indian Railways. Integrating NLP to Railways databases, optimizing their data retrieval processes using NL to SQL conversions, and database performance .
DIARY LOGS & CONNECTIONS
Personal diary writing (neurodiversity and system architecture) and connect portals (GitHub, LinkedIn, YouTube vlogs).
SYS_ADMIN PROFILE // LILYCORE
hi, welcome to my tiny space on the internet. i'm lily(ishita), an ai engineer currently micro-managing my neurodivergent cognitive runtime with transistors, binaries, and adequate amounts of caffeine. i build systems that run fast, look raw, and work elegantly with a sense of art (cus what is humanity without ARTS).
i have too many interests, too many wishes, and too many tasks i handle at once and somehow it all makes sense and works out perfectly?? i am a religious planner and want a sound structure to everything their is in my life. in this household, we crash out but NEVER BACK DOWN. nope, nada. its gonna sound corny, but how will you beat someone who lives for experiences and genuinely enjoys their work? (lol, im not talking about those evil antagonists who revel in people's misery.)
- [ JOURNAL ] i journal a lott.
- [ GUITAR ] playing guitar and singing occasionally.
- [ RUNNING ] trying to run almost everyday (i have a chronic fear about zombies running after me that haunts my dreams).
- [ TALKING ] absolute prodigy in yapping about anything and everything (trust me bro)
- [ POLYGLOT ] can speak french, korean and japanese(learning nihongo) i looove learning about cultures.
- [ MOVIES ] chronic movie watcher and a very easy-to-impress Letterboxd critique.
DOMAINS & IDEOLOGY
"Ever since I was eight, watching “How the Universe Works” on Discovery Science after school, I knew I wanted to be a scientist. I wanted to build something meaningful, something valuable that helps people or our resilient Earth."
My engineering ideology is deeply inspired by Eco-Brutalist Architecture. I believe that software should expose its structural mechanics, layout, and interfaces with raw honesty. Today's web is bloated with cosmetic animations, glass-morphs, tracking scripts, and complex abstraction layers that mask fragile foundations. I am weary of vibe coders replicating identical templates for their AI-wrapper startups.
Becoming an AI engineer while simultaneously being an AI critic may sound contradictory. However, consider this: I refuse to believe that Alan Turing, in all his brilliance, intended for AI to stifle creativity and encourage the use of GenAI for trivial tasks like designing flyers or altering photos, which can be misused or, even worse, employed as precise-target zones for warfare (those familiar with Turing would recognize the irony of this compared to his true vision). The pioneers of AI envisioned it predicting cancer six years ago, identifying regions on Earth more susceptible to droughts, and utilizing it for sustainable technology. It is my responsibility, as a daughter of Science, to contribute something—anything, in fact—toward building. Every engineer was born to build. I priorritise building systems that:
- Respect constraints: Work elegantly within resources rather than demanding more compute.
- Expose the plumbing: Design APIs and interfaces that don't hide their modes of failure.
- Resist decay: Write clean, dependency-minimal code that stands the test of time.
Click any research domain or core tenet above to extract system payload...
CURRENT WORK // CRIS DIVISION
I currently work at the Centre for Railway Information Systems (CRIS), an autonomous organization under the Ministry of Railways, India. CRIS builds and maintains the core IT backbone of Indian Railways.
Working on one of the world's largest rail networks comes with high-concurrency systems constraints:
- Real-Time Dynamic Data: Optimizing data retrieval of thousands of tables.
- Database performance: Tuning relational schema architectures, indexes, and caching strategies to process millions of train scheduling events, status logs, and booking records daily.
- High Accountabilty & Accuracy of Chatbot: Ensured robust fallback mechanisms like AST parsers and feedback-loop agents to keep the retrieved data in check.
PROJECTS ARCHIVE
SAFE_MED_RL
A safe Reinforcement Learning framework for controlling safety-critical medical devices. Implements DQN and PPO agents constrained within safety envelopes to prevent toxic dosing or system damage.
SECURE_NL2SQL
Natural Language to SQL converter. Translates complex human queries into SQL while piping syntax trees through rigorous defensive lint rules, semantic sandboxing, and SQL injection prevention systems.
> SECURITY_ALERT: DESTRUCTIVE ACTION DETECTED
> STATE: QUERY_REJECTED
CONTROLLER_BENCHMARK
Real-time analysis suite comparing traditional control loop controllers (PID) with deep learning architectures (PPO, DQN) in simulated chemical process plants. Exposes latency, overshoot parameters, and hardware cycles.
CERN_L1_ANOMALY_TRIGGER
An ultra-low latency, real-time Level-1 trigger pipeline detecting transient astronomical anomalies in ZTF light curve data. Compiles a lightweight Autoencoder (151 parameters) to synthesizable FPGA C++ firmware using hls4ml, achieving sub-100ns latency at 200MHz.
> Target Latency: < 100 ns // II = 1
> Precision Config: ap_fixed<16,6>
BLOGS & CONNECT
Establish data link with sys_admin:
System Diary Active
Select a diary log from the list above to mount its text blocks...