The Delhi junior engineer and the heritage-conservation bottleneck

His first three projects were straightforward: a 1980s RCC government office building in Okhla needing facade strengthening; a primary school in East Delhi requiring structural certification for RERA compliance. Done in eight weeks each. Approved by the supervising government engineer, invoiced, paid, filed away. The kind of work that teaches a young engineer that the system works — slowly, but works.

In May 2024, his firm won a tender for the structural retrofit of a 1920s haveli in Chandni Chowk. The building is five stories of load-bearing brick with lime-mortar joints, jali windows in cast-iron frames, a central courtyard with internal wells, stone flag flooring. It had not been substantially altered since 1928. The CPWD brief was straightforward: repair structural cracks, retrofit with hidden RCC bracing to meet 2002 seismic codes (the IS 1893 standard), preserve every visible surface, modernize mechanical and electrical systems without visible ducts or conduits. Simple, except that simplicity is not the word that applies when your material tolerance is the thickness of a pencil line and your client is the Heritage Conservation Committee.

He had three months to design. He took five, reworking every sheet: the reinforcement hidden behind the jali pattern; the moment connections placed between the innermost wythe of brick and a new internal frame; the mechanical riser concealed within a heritage-listed chimney stack (with approval to modify the exterior crown); the electrical conduit routed along the inside of the courtyard well, painted to match the 1928 plaster tone. He submitted on a Friday in October 2024, confident in the way young engineers are confident, which is entirely.

🗓️ The annual ritual

The Heritage Conservation Committee — a five-member panel of architects, conservation specialists, and government officials — exists to ensure that Old Delhi does not become a hollow reconstruction. The notified heritage zones cover 70+ precincts; 8,000+ properties fall under HCC oversight. The committee meets monthly. They have one rule that is written down and a dozen that are not: visible preservation is absolute; functional modernization is acceptable if it is invisible; and the correct balance between these two is a matter of educated judgment that varies by zone, by panel, and by which architect happens to be in the mood to defend the original 1920s construction logic.

The HCC design guidelines — a two-volume set printed by INTACH in 2008 and updated sporadically — specify that jali windows must retain original proportions, materials (cast iron, not steel), and operational mechanics. They do not specify what "original proportions" means when a jali has been replaced twice in the past thirty years and the original 1920s drawings have been lost. They do not specify what tolerance is acceptable when a retrofit seismic frame requires the jali to sit 30 mm further inboard than it did in 1928 — invisible to the street, but measurably different from the original.

Arjun's firm has learned to expect, on the first submission, one of three outcomes: approval (20% of cases); conditional approval pending minor revisions (40%); rejection with a request for "design re-conceptualization" (40%). Re-conceptualization is the phrase that means: the panel's intuition says something is wrong, and the panel would like you to guess what, and resubmit in six weeks.

The haveli submission was rejected in November 2024. The feedback, two pages typed single-spaced, noted that the proposed jali window details "did not adequately express the original 1920s aesthetic vocabulary" and suggested "reconsideration of the cast-iron proportions to better reflect neighboring properties in the zone." Arjun had spent seventy hours on the jali detailing alone. He called his father. His father said, quietly, that this was normal, and that he should redraw everything and resubmit.

He did. Second submission in January 2025. Rejected again. This time the feedback mentioned the courtyard well-routing for electrical conduit, and asked whether the conduit path was "architecturally justified" — a term that does not appear in the IS 1893 seismic code or the CPWD structural brief. Architecturally justified. He called his father again. His father said that this was also normal, and that he should consider hiring a heritage architect to collaborate.

Arjun's firm did. Third submission in March 2025, now with a Delhi Centre for Heritage architect credited. Rejected. The panel noted that while the seismic retrofit strategy was "technically sound," the integration of mechanical systems was still under-developed and required "further contextual study of the pre-1950 construction logic of the precinct."

It was March. It was supposed to be October.

What prevented the drowning was a Tuesday morning in late April when Vikrant handed Arjun a tablet and said that a colleague had shown him a new agent that does heritage-committee work, and that there was no harm in asking it to look at the four rejections and see if it could find a pattern.

🌗 What changed

The agent was loaded with Urdu and Hindi language packs, and Arjun, without much hope, opened his laptop and fed it the four HCC feedback letters. He asked, in Hindi, a question he had been asking for nine months: "मेरे चार rejection letters को देखो। क्या कोई pattern है कि HCC को क्या problem है? क्या मैं अगली submission में कोई चीज़ बेहतर कर सकता हूँ?" — Look at my four rejection letters. Is there a pattern to what HCC's problem is? Can I improve anything in the next submission?

The agent read the feedback letters in sequence and produced a document that changed the trajectory of the project.

"अरुण भैया, मैंने सभी चार rejections को analyze किया। तीन चीज़ें clear हो रही हैं: एक, HCC को jali windows के 'proportions' की definition नहीं है, पर उन्हें lagta है कि तुम्हारे नए proportions पुरानी हवेली के साथ match नहीं कर रहे हैं। दो, electrical-conduit routing के लिए 'architectural justification' यानी की: क्या उस routing को 1920s के construction logic के हिसाब से जस्टिफाई कर सकते हो — सिर्फ code से नहीं, बल्कि building की अपनी logic से। तीन, HCC को 'pre-1950 construction logic' की phrase से पता चल रहा है कि वो चाहते हैं कि तुम Chandni Chowk के पास-पड़ोस की दूसरी हवेलियों के सेम period की construction techniques को समझो, और अपने retrofit को उस context में explain करो।"

(Arjun bhai, I have analyzed all four rejections. Three things are clear: one, HCC has no definition of what jali window "proportions" should be, but they feel your new proportions do not match the original haveli. Two, the electrical-conduit routing needs "architectural justification" — that means: can you justify that routing based on the 1920s construction logic of the building itself, not just code compliance. Three, when HCC mentions "pre-1950 construction logic," they are telling you they want you to study the neighboring havelis from the same period in Chandni Chowk, understand their construction techniques, and explain your retrofit in that context.)

Arjun read it three times. The agent was not giving him the answer. It was giving him the question beneath the question.

The agent then did something that saved six weeks. It cross-checked the four HCC feedback letters against a database of 43 approved heritage-retrofit projects in Delhi completed between 2018 and 2025 — projects that had passed the same five-person committee or its predecessor members. It produced a comparative matrix: jali-window reinterpretations that the committee had approved in three similar 1920s havelis in nearby Karim Bazaar and Khari Baoli; mechanical-conduit routing strategies that had been accepted in three pre-1950 buildings in Shahjahanabad; and a list of which panel members (cross-referenced by name to the public HCC roster) had approved those projects.

"देखो, ये तीन havelis में (1) jali windows को 30-40 mm inboard रखा गया था, एक ही वजह से जिस वजह से तुमको करना है — RCC frame के लिए clear space चाहिए। जो architect Karim Bazaar की haveli को approve किया, वही दिल्ली का सबसे सीनियर heritage conservator है। उसका नाम Arun Srivastava है। अगर तुम अपनी jali detailing को 'structural-frame integration के लिए unavoidable horizontal offset' के रूप में present करो — सिर्फ 'proportions' की बजाय — और Srivastava के 2020 project की parallel लगा दो, तो committee को एक precedent मिल जाएगा।"

(Look — in three havelis here, (1) the jali windows were set 30-40 mm inboard for the exact reason you need to do it — RCC frame requires clear space. The architect who approved the Karim Bazaar haveli is Delhi's most senior heritage conservator. His name is Arun Srivastava. If you present your jali detailing not as a question of "proportions," but as an "unavoidable horizontal offset for structural-frame integration," and reference Srivastava's 2020 project as parallel, the committee gets a precedent.)

The agent also clarified the ambiguity of "pre-1950 construction logic." It was not vague philosophy; it was a coded signal that the committee wanted to see evidence of: Why was the electrical conduit routed through the courtyard well? Because in 1920, the well was the building's only interior void shaft, and the haveli's original electrification (added in 1950) had done the same thing. The retrofit was not imposing modern logic; it was following the building's own precedent. The agent found the 1950 electrical installation drawings (archived in the Delhi Municipal Corporation records database, accessible via the CPWD document-retrieval system) and placed them beside Arjun's design. The comparison took forty minutes to assemble and answered a nine-month question in photographic evidence.

Arjun revised the submission once more — fourth and, as it turned out, final. He re-framed the jali detailing as "structural-integration offset" with a precedent reference to Srivastava's 2020 Karim Bazaar project. He explained the electrical routing with the 1950 photographs and a three-sentence note: "The 1920s haveli lacked electrification. In 1950, the building was modernized by routing new conduit through the courtyard well, a vertical void — the building's pre-existing interior circulation spine. The retrofit follows the building's own logic rather than imposing new logic." He included a two-page "Contextual Analysis of Pre-1950 Construction Logic in Chandni Chowk Havelis" with photographs of three neighboring buildings' mechanical routing and a short note: in all cases, the voids and walls of the 1920s design had become the path of least resistance for 1950–1975 modernizations.

He submitted in early May 2025. The HCC met three weeks later. Approval came on a Tuesday, with a single line of feedback: "Design integrates seismic requirements within the architectural and historical logic of the building. Proceed to construction."

Arjun printed the approval letter, framed it, and hung it on the wall of his desk. It cost nothing. It had taken nine months.

:::infographic type: compare caption: "Three design strategies that the agent surfaced — each of which had already been approved by the HCC in prior projects." alt: "Comparison of jali-window detailing, electrical routing, and contextual approach in approved heritage projects." columns:

• icon: "🗽" title: "Jali-window offset (structural)" badge: "Precedent: Srivastava, Karim Bazaar 2020" body: "HCC had approved a 35 mm inboard offset in a geometrically similar 1920s haveli, justified as structural-frame integration. By referencing that approval, Arjun's offset — 32 mm — had a pre-existing defense."
• icon: "⚡" title: "Electrical conduit in courtyard well" badge: "Precedent: Original 1950 retrofit" body: "When the haveli was electrified in 1950, the same well shaft was used as the conduit path. The retrofit was following the building's own historical logic, not imposing new logic. Documentary proof existed in municipal archives."
• icon: "🏛️" title: "Contextual construction analysis" badge: "Three neighboring havelis, same period" body: "All three neighboring 1920s havelis had routed mechanical systems through pre-existing voids (wells, chimney chases, wall cavities) rather than through new penetrations. The pattern was consistent across the Chandni Chowk precinct."

🧭 Why we built it

There are approximately 85,000 licensed engineers in Delhi — 5% of India's total. Among them, roughly 1,400 work in heritage conservation or are learning it on the job, as Arjun was. Most are young. Most have come through one of five engineering schools — IIT Delhi, Delhi Technological University, TERI, Jamia Millia, TERI School of Advanced Studies — where the curriculum teaches codes and load paths but does not teach that a Heritage Conservation Committee is not a committee that can be understood by reading. It is a committee that understands itself by remembering the last seventy buildings it approved and rejected, a memory that is not written down and is not equally distributed across its five members.

For Arjun, what was missing was not engineering skill. It was precedent access. Every resubmission would have succeeded in 2–4 weeks if he had known, at the first rejection, that Srivastava had already approved a similar offset five years earlier. Every rejection letter that referenced "architectural justification" would have been answered in days if he had known that the 1950 electrical routing existed and was accessible in the municipal records. The HCC was not wrong. It was not irrational. It was asking him to prove his design in a language that requires fluency in forty years of committee decisions.

What the system does not provide is that fluency. The CPWD publishes empanelment criteria. DelhiRERA publishes compliance checklists. The Heritage Conservation Committee publishes a design guideline that is philosophically sound but not operationally specific. A young engineer inherits a first heritage project without a map of what has already been approved, no way to know that a design decision that feels novel to them is actually precedented, no archive of the committee's intuitions translated into pattern.

:::infographic type: boundary caption: "The boundary of what the agent does and does not do in heritage-approval cycles." alt: "Comparison of agent capabilities and limitations in HCC design work." does:

• icon: "🔍" text: "Analyzes rejection feedback to surface the implicit question underneath the coded language — turning 'architectural justification' into 'explain this choice using the building's own logic, not just code.'"
• icon: "📊" text: "Cross-checks proposed design details against approved HCC projects (2018–2025) and surfaces precedent when one exists — allowing designers to reference a real approval rather than guess."
• icon: "🗂️" text: "Retrieves archival documents (1950 electrical installation drawings, municipal records, prior renovation photographs) that prove a design choice follows the building's own history rather than imposing new logic." doesnt:
• icon: "🔒" text: "Never submits a design to HCC, never calls the committee, never attempts to influence the panel — the design and the submission strategy remain entirely the designer's decision."
• icon: "📝" text: "Never writes the response letter or the resubmission narrative — it surfaces patterns and precedent; the designer owns the argument and the framing."
• icon: "💭" text: "Never guarantees approval — HCC feedback remains subjective and panel-dependent. What the agent does is reduce the number of resubmissions needed to surface what the panel actually wants."

We built it because heritage conservation in India is not a separate profession. It is a tax on young engineers. A tax that is not well-calibrated to the person paying it. An IIT engineer with a seismic-retrofit brief should not need to spend nine months guessing at a committee's aesthetic intuitions, nor should they need to hire a separate heritage architect because the feedback is coded in a language that is not taught in the curriculum.

The precedent access changes the shape of the problem. It does not make the problem disappear. The committee will still reject designs that do not fit the precinct. But a rejection, now, will come with visibility: this design differs from the ones we approved in these three specific ways — rather than the nine-month version, which is this does not feel right, please reconsider.

🌱 What we hope happens

Three months after the approval, Arjun was back at his desk, now designing a second heritage-retrofit project — a 1960s government building in the Nizamuddin heritage precinct that required both seismic upgrade and facade restoration. The second project was faster: twelve weeks from submission to approval. The CPWD supervising engineer asked him, lightly, why the second project had moved so quickly. Arjun said, honestly, that he had learned something from the first one.

What we hope is that a young engineer in Delhi does not need to learn heritage conservation through nine months of rejection and rework. That a first project is hard because heritage design is hard — because the trade-offs between original form and modern safety are real and require judgment — but not hard because the decision criteria are opaque and the success cases are invisible.

Arjun's firm still employs a heritage architect. But now that architect is a collaborator, not a rescuer brought in after four rejections. The precedent database lives on a tablet in the office. When a new project lands, the team pulls up approved projects from the same precinct, the same period, the same committee members. The pattern becomes visible. The next resubmission, if needed, will come with a precedent reference. And the one after that will come faster still.

The HCC's intuitions are real. They are built on forty years of Delhi's heritage being carefully stewarded, one building at a time. They are not written in a ledger. They live in the memory of five people. Arjun's job — and the job of every structural engineer who works in a heritage zone — should be to access that memory, understand it, and engage it as a peer.

We built the agent so that Arjun could do that in his first month, rather than his ninth.